setiastrosuitepro 1.6.1__py3-none-any.whl

setiastro/saspro/live_stacking.py

@@ -0,0 +1,1830 @@
+from __future__ import annotations
+import os
+import glob
+import shutil
+import tempfile
+import datetime as _dt
+import numpy as np
+import time
+
+from PyQt6.QtCore import Qt, QTimer, QSettings, pyqtSignal
+from PyQt6.QtGui import QIcon, QImage, QPixmap, QAction, QIntValidator, QDoubleValidator
+from PyQt6.QtWidgets import (QDialog, QWidget, QLabel, QPushButton, QVBoxLayout, QHBoxLayout, QLineEdit,
+                             QFormLayout, QDialogButtonBox, QToolBar, QToolButton, QFileDialog,
+                             QSizePolicy, QGraphicsView, QGraphicsScene, QGraphicsPixmapItem, QApplication,
+                             QMessageBox, QSlider, QCheckBox, QInputDialog, QComboBox)
+
+import pyqtgraph as pg
+from astropy.io import fits
+from astropy.stats import sigma_clipped_stats
+
+# optional deps used in your code; guard if not installed
+try:
+    import rawpy
+except Exception:
+    rawpy = None
+try:
+    import exifread
+except Exception:
+    exifread = None
+
+import sep
+import exifread
+
+# your helpers/utilities
+from setiastro.saspro.imageops.stretch import stretch_mono_image, stretch_color_image
+from setiastro.saspro.legacy.numba_utils import apply_flat_division_numba, debayer_fits_fast   # adjust names if different
+from setiastro.saspro.legacy.image_manager import load_image
+from setiastro.saspro.star_alignment import StarRegistrationWorker, StarRegistrationThread, IDENTITY_2x3
+from setiastro.saspro.widgets.spinboxes import CustomSpinBox, CustomDoubleSpinBox
+
+
+class LiveStackSettingsDialog(QDialog):
+    """
+    Combined dialog for:
+      • Live‐stack parameters (bootstrap frames, σ‐clip threshold)
+      • Culling thresholds (max FWHM, max eccentricity, min star count)
+    """
+    def __init__(self, parent):
+        super().__init__(parent)
+        self.setWindowTitle("Live Stack & Culling Settings")
+
+        # — Live Stack Settings —
+        # Bootstrap frames (int)
+        self.bs_spin = CustomSpinBox(
+            minimum=1,
+            maximum=100,
+            initial=parent.bootstrap_frames,
+            step=1
+        )
+        self.bs_spin.valueChanged.connect(lambda v: None)
+
+        # Sigma threshold (float)
+        self.sigma_spin = CustomDoubleSpinBox(
+            minimum=0.1,
+            maximum=10.0,
+            initial=parent.clip_threshold,
+            step=0.1
+        )
+        self.sigma_spin.valueChanged.connect(lambda v: None)
+
+        # — Culling Thresholds —
+        # Max FWHM (float)
+        self.fwhm_spin = CustomDoubleSpinBox(
+            minimum=0.1,
+            maximum=50.0,
+            initial=parent.max_fwhm,
+            step=0.1
+        )
+        self.fwhm_spin.valueChanged.connect(lambda v: None)
+
+        # Max eccentricity (float)
+        self.ecc_spin = CustomDoubleSpinBox(
+            minimum=0.0,
+            maximum=1.0,
+            initial=parent.max_ecc,
+            step=0.01
+        )
+        self.ecc_spin.valueChanged.connect(lambda v: None)
+
+        # Min star count (int)
+        self.star_spin = CustomSpinBox(
+            minimum=0,
+            maximum=5000,
+            initial=parent.min_star_count,
+            step=1
+        )
+        self.star_spin.valueChanged.connect(lambda v: None)
+        
+        # Acquisition Dely (int)
+        self.delay_spin = CustomDoubleSpinBox(
+            minimum=0.0,
+            maximum=60.0,
+            initial=parent.FILE_STABLE_SECS,
+            step=0.5
+        )
+
+
+
+        # Build form layout
+        form = QFormLayout()
+        form.addRow("Switch to μ–σ clipping after:", self.bs_spin)
+        form.addRow("Clip threshold:", self.sigma_spin)
+        form.addRow(QLabel(""))  # blank row for separation
+        form.addRow("Max FWHM (px):", self.fwhm_spin)
+        form.addRow("Max Eccentricity:", self.ecc_spin)
+        form.addRow("Min Star Count:", self.star_spin)
+        form.addRow("Acquisition Delay:", self.delay_spin)
+
+        self.mapping_combo = QComboBox()
+        opts = ["Natural", "SHO", "HSO", "OSH", "SOH", "HOS", "OHS"]
+        self.mapping_combo.addItems(opts)
+        # preselect current
+        idx = opts.index(parent.narrowband_mapping) \
+              if parent.narrowband_mapping in opts else 0
+        self.mapping_combo.setCurrentIndex(idx)
+        form.addRow("Narrowband Mapping:", self.mapping_combo)
+
+        # OK / Cancel buttons
+        btns = QDialogButtonBox(
+            QDialogButtonBox.StandardButton.Ok | QDialogButtonBox.StandardButton.Cancel
+        )
+        btns.accepted.connect(self.accept)
+        btns.rejected.connect(self.reject)
+
+        # Assemble dialog layout
+        layout = QVBoxLayout()
+        layout.addLayout(form)
+        layout.addWidget(btns)
+        self.setLayout(layout)
+
+    def getValues(self):
+        """
+        Returns a tuple of five values in order:
+          (bootstrap_frames, clip_threshold,
+           max_fwhm, max_ecc, min_star_count, delay)
+        """
+        bs      = self.bs_spin.value
+        sigma   = self.sigma_spin.value()
+        fwhm    = self.fwhm_spin.value()
+        ecc     = self.ecc_spin.value()
+        stars   = self.star_spin.value
+        mapping = self.mapping_combo.currentText()
+        delay = self.delay_spin.value()
+        return bs, sigma, fwhm, ecc, stars, mapping, delay
+
+
+
+class LiveMetricsPanel(QWidget):
+    """
+    A simple 2×2 grid of PyQtGraph plots to show, in real time:
+      [0,0] → FWHM (px) vs. frame index
+      [0,1] → Eccentricity vs. frame index
+      [1,0] → Star Count vs. frame index
+      [1,1] → (μ–ν)/σ (∝SNR) vs. frame index
+    """
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        titles = ["FWHM (px)", "Eccentricity", "Star Count", "(μ–ν)/σ (∝SNR)"]
+
+        layout = QVBoxLayout(self)
+        grid = pg.GraphicsLayoutWidget()
+        layout.addWidget(grid)
+
+        self.plots = []
+        self.scats = []
+        self._data_x = [[], [], [], []]
+        self._data_y = [[], [], [], []]
+        self._flags  = [[], [], [], []]  # track if each point was “bad” (True) or “good” (False)
+
+        for row in range(2):
+            for col in range(2):
+                pw = grid.addPlot(row=row, col=col)
+                idx = row * 2 + col
+                pw.setTitle(titles[idx])
+                pw.showGrid(x=True, y=True, alpha=0.3)
+                pw.setLabel('bottom', "Frame #")
+                pw.setLabel('left', titles[idx])
+
+                scat = pg.ScatterPlotItem(pen=pg.mkPen(None),
+                                          brush=pg.mkBrush(100, 100, 255, 200),
+                                          size=6)
+                pw.addItem(scat)
+                self.plots.append(pw)
+                self.scats.append(scat)
+
+    def add_point(self, frame_idx: int, fwhm: float, ecc: float, star_cnt: int, snr_val: float, flagged: bool):
+        """
+        Append one new data point to each metric.  
+        If flagged == True, draw that single point in red; else blue.  
+        But keep all previously-plotted points at their original colors.
+        """
+        values = [fwhm, ecc, star_cnt, snr_val]
+        for i in range(4):
+            self._data_x[i].append(frame_idx)
+            self._data_y[i].append(values[i])
+            self._flags[i].append(flagged)
+
+            # Now build a brush list for *all* points up to index i,
+            # coloring each point according to its own flag.
+            brushes = [
+                pg.mkBrush(255, 0, 0, 200) if self._flags[i][j]
+                else pg.mkBrush(100, 100, 255, 200)
+                for j in range(len(self._data_x[i]))
+            ]
+
+            self.scats[i].setData(
+                self._data_x[i],
+                self._data_y[i],
+                brush=brushes,
+                pen=pg.mkPen(None),
+                size=6
+            )
+
+    def clear_all(self):
+        """Clear data from all four plots."""
+        for i in range(4):
+            self._data_x[i].clear()
+            self._data_y[i].clear()
+            self._flags[i].clear()
+            self.scats[i].clear()
+
+class LiveMetricsWindow(QWidget):
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        self.setWindowTitle("Live Stack Metrics")
+        self.resize(600, 400)
+
+        layout = QVBoxLayout(self)
+        self.metrics_panel = LiveMetricsPanel(self)
+        layout.addWidget(self.metrics_panel)
+
+from setiastro.saspro.star_metrics import measure_stars_sep
+
+def compute_frame_star_metrics(image_2d):
+    """
+    Harmonized with Blink metrics:
+      - SEP.Background() for back/noise
+      - thresh = 7σ
+      - median aggregation for FWHM & Ecc
+    """
+    # ensure float32 mono [0..1]
+    data = np.asarray(image_2d)
+    if data.ndim == 3:
+        data = data.mean(axis=2)
+    if data.dtype == np.uint8:
+        data = data.astype(np.float32) / 255.0
+    elif data.dtype == np.uint16:
+        data = data.astype(np.float32) / 65535.0
+    else:
+        data = data.astype(np.float32, copy=False)
+
+    star_count, fwhm, ecc = measure_stars_sep(
+        data,
+        thresh_sigma=7.0,
+        minarea=16,
+        deblend_nthresh=32,
+        aggregate="median",
+    )
+    return star_count, fwhm, ecc
+
+def estimate_global_snr(
+    stack_image: np.ndarray,
+    bkg_box_size: int = 200
+) -> float:
+    """
+    “Hybrid” global SNR ≔ (μ_patch − median_patch) / σ_central,
+    where:
+      • μ_patch    and median_patch come from a small bkg_box_size×bkg_box_size patch
+        centered inside the middle 50% of the image.
+      • σ_central  is the standard deviation computed over the entire “middle 50%” region.
+
+    Steps:
+      1) Collapse to grayscale (H×W) if needed.
+      2) Identify the middle 50% rectangle of the image.
+      3) Within that, center a patch of size up to bkg_box_size×bkg_box_size.
+      4) Compute μ_patch = mean(patch), median_patch = median(patch).
+      5) Compute σ_central = std(middle50_region).
+      6) If σ_central ≤ 0, return 0. Otherwise return (μ_patch − median_patch) / σ_central.
+    """
+
+    # 1) Collapse to simple 2D float array (grayscale)
+    if stack_image.ndim == 3 and stack_image.shape[2] == 3:
+        # RGB → grayscale by averaging channels
+        gray = stack_image.mean(axis=2).astype(np.float32)
+    else:
+        # Already mono: just cast to float32
+        gray = stack_image.astype(np.float32)
+
+    H, W = gray.shape
+
+    # 2) Compute coordinates of the “middle 50%” rectangle
+    y0 = H // 4
+    y1 = y0 + (H // 2)
+    x0 = W // 4
+    x1 = x0 + (W // 2)
+
+    # Extract that central50 region as a view (no copy)
+    central50 = gray[y0:y1, x0:x1]
+
+    # 3) Within that central50, choose a patch of up to bkg_box_size×bkg_box_size, centered
+    center_h = (y1 - y0)
+    center_w = (x1 - x0)
+
+    # Clamp box size so it does not exceed central50 dimensions
+    box_h = min(bkg_box_size, center_h)
+    box_w = min(bkg_box_size, center_w)
+
+    # Compute top-left corner of that patch so it’s centered in central50
+    cy0 = y0 + (center_h - box_h) // 2
+    cx0 = x0 + (center_w - box_w) // 2
+
+    patch = gray[cy0 : cy0 + box_h, cx0 : cx0 + box_w]
+
+    # 4) Compute patch statistics
+    mu_patch = float(np.mean(patch))
+    med_patch = float(np.median(patch))
+    min_patch = float(np.min(patch))
+
+    # 5) Compute σ over the entire central50 region
+    sigma_central = float(np.std(central50))
+    if sigma_central <= 0.0:
+        return 0.0
+
+    nu = med_patch - 3.0 * sigma_central * med_patch
+
+    # 6) Return (mean − nu) / σ
+    return (mu_patch - nu) / sigma_central
+    #return (mu_patch) / sigma_central
+
+class LiveStackWindow(QDialog):
+    def __init__(self, parent=None, doc_manager=None, wrench_path=None, spinner_path=None):
+        super().__init__(parent)
+        self.parent = parent
+        self._docman = doc_manager
+        self._wrench_path = wrench_path
+        self._spinner_path = spinner_path
+        self.setWindowTitle("Live Stacking")
+        self.resize(900, 600)
+
+        # ─── State Variables ─────────────────────────────────────
+        self.watch_folder = None
+        self.processed_files = set()
+        self.master_dark = None
+        self.master_flat = None
+        self.master_flats  = {}
+
+        self.filter_stacks = {}       # key → np.ndarray (float32)
+        self.filter_counts = {}       # key → int
+        self.filter_buffers  = {}  # key → list of bootstrap frames [H×W arrays]
+        self.filter_mus      = {}  # key → µ array after bootstrap (H×W)
+        self.filter_m2s      = {}  # key → M2 array after bootstrap (H×W)
+
+        self.cull_folder = None
+
+        self.is_running = False
+        self.frame_count = 0
+        self.current_stack = None
+
+        self._probe = {}  # path -> {"size": int, "mtime": float, "since": float, "penalty_until": float}
+        # Tunables:
+        self.FILE_STABLE_SECS = 3.0          # how long size+mtime must stay unchanged
+        self.OPEN_RETRY_PENALTY_SECS = 10.0  # cool-down after a read/permission failure
+        self.MAX_FILE_WAIT_SECS = 600.0      # optional safety cap (unused by default logic)
+
+        # ── Load persisted settings ───────────────────────────────
+        s = QSettings()
+        self.bootstrap_frames    = s.value("LiveStack/bootstrap_frames",    24,     type=int)
+        self.clip_threshold      = s.value("LiveStack/clip_threshold",      3.5,    type=float)
+        self.max_fwhm            = s.value("LiveStack/max_fwhm",            15.0,   type=float)
+        self.max_ecc             = s.value("LiveStack/max_ecc",             0.9,    type=float)
+        self.min_star_count      = s.value("LiveStack/min_star_count",      5,      type=int)
+        self.narrowband_mapping  = s.value("LiveStack/narrowband_mapping",  "Natural", type=str)
+        self.star_trail_mode = s.value("LiveStack/star_trail_mode", False, type=bool)
+
+
+        self.total_exposure = 0.0  # seconds
+        self.exposure_label = QLabel("Total Exp: 00:00:00")
+        self.exposure_label.setStyleSheet("color: #cccccc; font-weight: bold;")
+
+        self.brightness = 0.0   # [-1.0..+1.0]
+        self.contrast   = 1.0   # [0.1..3.0]
+
+
+        self._buffer = []    # store up to bootstrap_frames normalized frames
+        self._mu = None      # per-pixel mean (after bootstrap)
+        self._m2 = None      # per-pixel sum of squares differences (for Welford)
+
+        # ─── Create Separate Metrics Window (initially hidden) ─────
+        # We do NOT embed this in the stacking dialog’s layout!
+        self.metrics_window = LiveMetricsWindow(None)
+        self.metrics_window.hide()
+
+        # ─── UI ELEMENTS FOR STACKING DIALOG ───────────────────────
+        # 1) Folder selection
+        self.folder_label = QLabel("Folder: (none)")
+        self.select_folder_btn = QPushButton("Select Folder…")
+        self.select_folder_btn.clicked.connect(self.select_folder)
+
+        # 2) Load master dark/flat
+        self.load_darks_btn = QPushButton("Load Master Dark…")
+        self.load_darks_btn.clicked.connect(self.load_masters)
+        self.load_flats_btn = QPushButton("Load Master Flat…")
+        self.load_flats_btn.clicked.connect(self.load_masters)
+        self.load_filter_flats_btn = QPushButton("Load MonoFilter Flats…")
+        self.load_filter_flats_btn.clicked.connect(self.load_filter_flats)        
+
+        # 2b) Cull folder selection
+        self.cull_folder_label = QLabel("Cull Folder: (none)")
+        self.select_cull_btn = QPushButton("Select Cull Folder…")
+        self.select_cull_btn.clicked.connect(self.select_cull_folder)
+
+        self.dark_status_label = QLabel("Dark: ❌")
+        self.flat_status_label = QLabel("Flat: ❌")
+        for lbl in (self.dark_status_label, self.flat_status_label):
+            lbl.setStyleSheet("color: #cccccc; font-weight: bold;")
+        # 3) “Process & Monitor” / “Monitor Only” / “Stop” / “Reset”
+        self.mono_color_checkbox = QCheckBox("Mono → Color Stacking")
+        self.mono_color_checkbox.setToolTip(
+            "When checked, bucket mono frames by FILTER and composite R, G, B, Ha, OIII, SII."
+        )
+        # **Connect the toggled(bool) signal** before we ever call it
+        self.mono_color_checkbox.toggled.connect(self._on_mono_color_toggled)
+
+        # ** new: Star-Trail mode checkbox **
+        self.star_trail_checkbox = QCheckBox("★★ Star-Trail Mode ★★")
+        self.star_trail_checkbox.setChecked(self.star_trail_mode)
+        self.star_trail_checkbox.setToolTip("If checked, build a max-value trail instead of a running stack")
+        self.star_trail_checkbox.toggled.connect(self._on_star_trail_toggled)
+
+        self.process_and_monitor_btn = QPushButton("Process && Monitor")
+        self.process_and_monitor_btn.clicked.connect(self.start_and_process)
+        self.monitor_only_btn = QPushButton("Monitor Only")
+        self.monitor_only_btn.clicked.connect(self.start_monitor_only)
+        self.stop_btn = QPushButton("Stop")
+        self.stop_btn.clicked.connect(self.stop_live)
+        self.reset_btn = QPushButton("Reset")
+        self.reset_btn.clicked.connect(self.reset_live)
+
+        self.frame_count_label = QLabel("Frames: 0")
+
+        # 4) Live‐stack preview area (QGraphicsView)
+        self.scene = QGraphicsScene(self)
+        self.view = QGraphicsView(self.scene, self)
+        self.view.setDragMode(QGraphicsView.DragMode.ScrollHandDrag)
+        self.view.setTransformationAnchor(QGraphicsView.ViewportAnchor.AnchorUnderMouse)
+        self.pixmap_item = QGraphicsPixmapItem()
+        self.scene.addItem(self.pixmap_item)
+        self._did_initial_fit = False
+
+        # 5) Zoom toolbar + Settings icon
+        tb = QToolBar()
+
+        zi = QAction(QIcon.fromTheme("zoom-in"), "Zoom In", self)
+        zo = QAction(QIcon.fromTheme("zoom-out"), "Zoom Out", self)
+        fit = QAction(QIcon.fromTheme("zoom-fit-best"), "Fit to Window", self)
+
+        tb.addAction(zi)
+        tb.addAction(zo)
+        tb.addAction(fit)
+
+        spacer = QWidget()
+        spacer.setSizePolicy(QSizePolicy.Policy.Expanding, QSizePolicy.Policy.Preferred)
+        tb.addWidget(spacer)
+        # — Replace the QAction “wrench” with a styled QToolButton —
+        self.wrench_button = QToolButton()
+        self.wrench_button.setIcon(QIcon(self._wrench_path))
+        self.wrench_button.setToolTip("Settings")
+        # Apply your stylesheet to the QToolButton
+        self.wrench_button.setStyleSheet("""
+            QToolButton {
+                background-color: #FF4500;
+                color: white;
+                font-size: 16px;
+                padding: 8px;
+                border-radius: 5px;
+                font-weight: bold;
+            }
+            QToolButton:hover {
+                background-color: #FF6347;
+            }
+        """)
+        # Connect the clicked signal to open_settings()
+        self.wrench_button.clicked.connect(self.open_settings)
+
+        # Add the styled QToolButton into the toolbar
+        tb.addWidget(self.wrench_button)
+
+        zi.triggered.connect(self.zoom_in)
+        zo.triggered.connect(self.zoom_out)
+        fit.triggered.connect(self.fit_to_window)
+
+
+        # 6) Brightness & Contrast sliders
+        bright_slider = QSlider(Qt.Orientation.Horizontal)
+        bright_slider.setRange(-100, 100)
+        bright_slider.setValue(0)
+        bright_slider.setToolTip("Brightness")
+        bright_slider.valueChanged.connect(self.on_brightness_changed)
+
+        contrast_slider = QSlider(Qt.Orientation.Horizontal)
+        contrast_slider.setRange(10, 1000)
+        contrast_slider.setValue(100)
+        contrast_slider.setToolTip("Contrast")
+        contrast_slider.valueChanged.connect(self.on_contrast_changed)
+
+        bc_layout = QHBoxLayout()
+        bc_layout.addWidget(QLabel("Brightness"))
+        bc_layout.addWidget(bright_slider)
+        bc_layout.addWidget(QLabel("Contrast"))
+        bc_layout.addWidget(contrast_slider)
+
+        # 7) “Send to Slot” button
+        open_btn = QPushButton("Open in New View ▶")
+        open_btn.clicked.connect(self.send_to_new_view)
+
+        # 8) “Show Metrics” button
+        self.show_metrics_btn = QPushButton("Show Metrics")
+        self.show_metrics_btn.clicked.connect(self.show_metrics_window)
+
+        # ─── ASSEMBLE MAIN LAYOUT (exactly one setLayout call!) ─────
+        main_layout = QVBoxLayout()
+
+        # A) Top‐row controls
+        controls = QHBoxLayout()
+        controls.addWidget(self.select_folder_btn)
+        controls.addWidget(self.load_darks_btn)
+        controls.addWidget(self.load_flats_btn)
+        controls.addWidget(self.load_filter_flats_btn)
+        controls.addWidget(self.select_cull_btn)
+        controls.addStretch()
+        controls.addWidget(self.mono_color_checkbox)
+        controls.addWidget(self.star_trail_checkbox)
+        controls.addWidget(self.process_and_monitor_btn)
+        controls.addWidget(self.monitor_only_btn)
+        controls.addWidget(self.stop_btn)
+        controls.addWidget(self.reset_btn)
+        main_layout.addLayout(controls)
+
+        # B) Status line: folder label + frame count
+        status_line = QHBoxLayout()
+        status_line.addWidget(self.folder_label)
+        status_line.addWidget(self.dark_status_label)
+        status_line.addWidget(self.flat_status_label)
+        status_line.addWidget(self.cull_folder_label)
+        status_line.addStretch()        
+        status_line.addWidget(self.frame_count_label)
+        status_line.addWidget(self.exposure_label)
+        main_layout.addLayout(status_line)
+
+        # C) Zoom toolbar
+        main_layout.addWidget(tb)
+
+        # D) Show Metrics button (separate window)
+        main_layout.addWidget(self.show_metrics_btn)
+
+        # E) Live‐stack preview area
+        main_layout.addWidget(self.view)
+
+        # F) Brightness/Contrast sliders
+        main_layout.addLayout(bc_layout)
+
+        # G) “Send to Slot” + mode/idle labels
+        main_layout.addWidget(open_btn)
+        self.mode_label = QLabel("Mode: Linear Average")
+        self.mode_label.setStyleSheet("color: #a0a0a0;")
+        main_layout.addWidget(self.mode_label)
+        self.status_label = QLabel("Idle")
+        self.status_label.setStyleSheet("color: #a0a0a0;")
+        main_layout.addWidget(self.status_label)
+
+        # Finalize
+        self.setLayout(main_layout)
+
+        # Timer for polling new files
+        self.poll_timer = QTimer(self)
+        self.poll_timer.setInterval(1500)
+        self.poll_timer.timeout.connect(self.check_for_new_frames)
+        self._on_mono_color_toggled(self.mono_color_checkbox.isChecked())
+
+
+    # ─────────────────────────────────────────────────────────────────────────
+    def _on_star_trail_toggled(self, checked: bool):
+        """Enable/disable star-trail mode."""
+        self.star_trail_mode = checked
+        QSettings().setValue("LiveStack/star_trail_mode", checked)
+        self.mode_label.setText("Mode: Star-Trail" if checked else "Mode: Linear Average")
+        # if you want, disable mono/color checkbox when star-trail is on:
+        self.mono_color_checkbox.setEnabled(not checked)
+
+    def _on_mono_color_toggled(self, checked: bool):
+        self.mono_color_mode = checked
+        self.filter_stacks.clear()
+        self.filter_counts.clear()
+
+        msg = "Enabled" if checked else "Disabled"
+        self.status_label.setText(f"Mono→Color Mode {msg}")
+
+    def show_metrics_window(self):
+        """Pop up the separate metrics window (never embed it here)."""
+        self.metrics_window.show()
+        self.metrics_window.raise_()
+
+
+    def select_cull_folder(self):
+        folder = QFileDialog.getExistingDirectory(self, "Select Cull Folder")
+        if folder:
+            self.cull_folder = folder
+            self.cull_folder_label.setText(f"Cull: {os.path.basename(folder)}")
+
+    def _cull_frame(self, path: str):
+        """
+        Move a flagged frame into the cull folder (if set), 
+        or just update the status label if not.
+        """
+        name = os.path.basename(path)
+        if self.cull_folder:
+            try:
+                os.makedirs(self.cull_folder, exist_ok=True)
+                dst = os.path.join(self.cull_folder, name)
+                shutil.move(path, dst)
+                self.status_label.setText(f"⚠ Culled {name} → {self.cull_folder}")
+            except Exception:
+                self.status_label.setText(f"⚠ Failed to cull {name}")
+        else:
+            self.status_label.setText(f"⚠ Flagged (not stacked): {name}")
+        QApplication.processEvents()
+
+    def open_settings(self):
+        dlg = LiveStackSettingsDialog(self)
+        if dlg.exec() == QDialog.DialogCode.Accepted:
+            bs, sigma, fwhm, ecc, stars, mapping, delay = dlg.getValues()
+
+            # 1) Persist into QSettings
+            s = QSettings()
+            s.setValue("LiveStack/bootstrap_frames",   bs)
+            s.setValue("LiveStack/clip_threshold",     sigma)
+            s.setValue("LiveStack/max_fwhm",           fwhm)
+            s.setValue("LiveStack/max_ecc",            ecc)
+            s.setValue("LiveStack/min_star_count",     stars)
+            s.setValue("LiveStack/narrowband_mapping", mapping)
+            s.setValue("LiveStack/file_stable_secs", delay)
+
+            # 2) Apply to this live‐stack session
+            self.bootstrap_frames   = bs
+            self.clip_threshold     = sigma
+            self.max_fwhm           = fwhm
+            self.max_ecc            = ecc
+            self.min_star_count     = stars
+            self.narrowband_mapping = mapping
+            self.FILE_STABLE_SECS = delay
+
+            self.status_label.setText(
+                f"↺ Settings saved: BS={bs}, σ={sigma:.1f}, "
+                f"FWHM≤{fwhm:.1f}, ECC≤{ecc:.2f}, Stars≥{stars}, "
+                f"Mapping={mapping}"
+            )
+            QApplication.processEvents()
+
+    def zoom_in(self):
+        self.view.scale(1.2, 1.2)
+
+    def zoom_out(self):
+        self.view.scale(1/1.2, 1/1.2)
+
+    def fit_to_window(self):
+        self.view.fitInView(self.scene.sceneRect(), Qt.AspectRatioMode.KeepAspectRatio)
+
+    # — Brightness / Contrast —
+
+    def _refresh_preview(self):
+        """
+        Recompute the current preview array (stack vs. composite)
+        and call update_preview on it.
+        """
+        if self.mono_color_mode:
+            # build the composite from filter_stacks
+            preview = self._build_color_composite()
+        else:
+            # use the normal running stack
+            preview = self.current_stack
+
+        if preview is not None:
+            self.update_preview(preview)
+
+    def on_brightness_changed(self, val: int):
+        self.brightness = val / 100.0  # map to [-1,1]
+        self._refresh_preview()
+
+    def on_contrast_changed(self, val: int):
+        self.contrast = val / 100.0   # map to [0.1,10.0]
+        self._refresh_preview()
+
+    # — Sending out —
+
+    def send_to_new_view(self):
+        """
+        Create a brand-new document/view from the current live stack or composite.
+        Prefers using doc_manager's native numpy-open methods; otherwise falls back
+        to writing a temp TIFF and asking the host window to open it.
+        """
+        # pick what to export
+        if self.mono_color_mode:
+            img = self._build_color_composite()
+        else:
+            img = self.current_stack
+
+        if img is None:
+            self.status_label.setText("⚠ Nothing to open")
+            return
+
+        # ensure float32 in [0,1]
+        img = np.clip(img, 0.0, 1.0).astype(np.float32)
+
+        title = f"LiveStack_{_dt.datetime.now():%Y%m%d_%H%M%S}_{self.frame_count}f"
+        metadata = {"source": "LiveStack", "frames_stacked": int(self.frame_count)}
+
+        # 1) Try doc_manager direct numpy APIs (several common names)
+        dm = self._docman
+        if dm is not None:
+            for name in ("create_numpy_document",
+                        "new_document_from_numpy",
+                        "open_numpy",
+                        "open_array",
+                        "open_image_array",
+                        "add_document_from_array"):
+                fn = getattr(dm, name, None)
+                if callable(fn):
+                    try:
+                        fn(img, title=title, metadata=metadata)
+                        self.status_label.setText(f"Opened new view: {title}")
+                        return
+                    except TypeError:
+                        # some variants might not accept title/metadata
+                        try:
+                            fn(img)
+                            self.status_label.setText(f"Opened new view: {title}")
+                            return
+                        except Exception:
+                            pass
+                    except Exception:
+                        pass
+
+        # 2) Fallback: write a temp 16-bit TIFF and ask main window to open it
+        try:
+            import tifffile as tiff
+            tmp = tempfile.NamedTemporaryFile(suffix=".tiff", delete=False)
+            tmp_path = tmp.name
+            tmp.close()
+            # export as 16-bit so it's friendly to the rest of the app
+            arr16 = np.clip(img * 65535.0, 0, 65535).astype(np.uint16)
+            tiff.imwrite(tmp_path, arr16)
+
+            host = self.parent
+            for name in ("open_files", "open_file", "load_paths", "load_path"):
+                fn = getattr(host, name, None)
+                if callable(fn):
+                    try:
+                        fn([tmp_path]) if fn.__code__.co_argcount != 2 else fn(tmp_path)
+                        self.status_label.setText(f"Opened new view from temp: {os.path.basename(tmp_path)}")
+                        return
+                    except Exception:
+                        pass
+
+            # ultimate fallback: let the user know where it went
+            QMessageBox.information(self, "Saved Temp Image",
+                                    f"Saved temp: {tmp_path}\nOpen it from File → Open.")
+        except Exception as e:
+            QMessageBox.warning(self, "Open Failed",
+                                f"Could not open in new view:\n{e}")
+
+
+    # ── New helper: map header["FILTER"] to a single letter key
+    def _get_filter_key(self, header):
+        """
+        Map a FITS header FILTER string to one of:
+        'L' (luminance),
+        'R','G','B',
+        'H' (H-alpha),
+        'O' (OIII),
+        'S' (SII),
+        or return None if it doesn’t match.
+        """
+        raw = header.get('FILTER', '')
+        fn = raw.strip().upper()
+        if not fn:
+            return None
+
+        # H-alpha
+        if fn in ('H', 'HA', 'HALPHA', 'H-ALPHA'):
+            return 'H'
+        # OIII
+        if fn in ('O', 'O3', 'OIII'):
+            return 'O'
+        # SII
+        if fn in ('S', 'S2', 'SII'):
+            return 'S'
+        # Red
+        if fn in ('R', 'RED', 'RD'):
+            return 'R'
+        # Green
+        if fn in ('G', 'GREEN', 'GRN'):
+            return 'G'
+        # Blue
+        if fn in ('B', 'BLUE', 'BL'):
+            return 'B'
+        # Luminance
+        if fn in ('L', 'LUM', 'LUMI', 'LUMINANCE'):
+            return 'L'
+
+        return None
+
+    # ── New helper: stack a single mono frame under filter key
+    def _stack_mono_channel(self, key, img, delta=None):
+        # img: 2D or 3D array; we convert to 2D mono always
+        mono = img if img.ndim==2 else np.mean(img,axis=2)
+        # align if you need (use same logic as color branch)
+        if hasattr(self, 'reference_image_2d'):
+            d = delta or StarRegistrationWorker.compute_affine_transform_astroalign(
+                        mono, self.reference_image_2d)
+            if d is not None:
+                mono = StarRegistrationThread.apply_affine_transform_static(mono, d)
+        # normalize
+        norm = stretch_mono_image(mono, target_median=0.3)
+        # first frame?
+        if key not in self.filter_stacks:
+            self.filter_stacks[key] = norm.copy()
+            self.filter_counts[key] = 1
+            # set reference on first good channel frame
+            if not hasattr(self, 'reference_image_2d'):
+                self.reference_image_2d = norm.copy()
+        else:
+            cnt = self.filter_counts[key]
+            self.filter_stacks[key] = (cnt/self.filter_counts[key]+1)*self.filter_stacks[key] \
+                                      + (1.0/(cnt+1))*norm
+            self.filter_counts[key] += 1
+
+    # ── New helper: build an RGB preview from whatever channels we have
+    def _build_color_composite(self):
+        """
+        Composite filters into an RGB preview according to self.narrowband_mapping:
+
+        • "Natural":
+            – If SII present:
+                R = 0.5*(Ha + SII)
+                G = 0.5*(SII + OIII)
+                B = OIII
+            – Elif any R/G/B loaded:
+                R = R_filter
+                G = G_filter + OIII
+                B = B_filter + OIII
+            – Else (no SII, no R/G/B):
+                R = Ha
+                G = OIII
+                B = OIII
+
+        • Any 3-letter code (e.g. "SHO", "OHS"):
+            R = filter_stacks[mapping[0]]
+            G = filter_stacks[mapping[1]]
+            B = filter_stacks[mapping[2]]
+
+        Missing channels default to zero.
+        """
+        # 1) Determine H, W
+        if self.filter_stacks:
+            first = next(iter(self.filter_stacks.values()))
+            H, W = first.shape
+        elif getattr(self, 'current_stack', None) is not None:
+            H, W = self.current_stack.shape[:2]
+        else:
+            return None
+
+        # helper: get stack or zeros
+        def getf(k):
+            return self.filter_stacks.get(k, np.zeros((H, W), np.float32))
+
+        mode = self.narrowband_mapping.upper()
+        if mode == "NATURAL":
+            Ha = getf('H')
+            O3 = getf('O')
+            S2 = self.filter_stacks.get('S', None)
+            Rf = self.filter_stacks.get('R', None)
+            Gf = self.filter_stacks.get('G', None)
+            Bf = self.filter_stacks.get('B', None)
+
+            if S2 is not None:
+                # narrowband SII branch
+                R = 0.5 * (Ha + S2)
+                G = 0.5 * (S2 + O3)
+                B = O3.copy()
+
+            elif any(x is not None for x in (Rf, Gf, Bf)):
+                # broadband branch: Rf/Gf/Bf with OIII boost
+                R = Rf if Rf is not None else np.zeros((H, W), np.float32)
+                G = (Gf if Gf is not None else np.zeros((H, W), np.float32)) + O3
+                B = (Bf if Bf is not None else np.zeros((H, W), np.float32)) + O3
+
+            else:
+                # fallback HOO
+                R = Ha
+                G = O3
+                B = O3
+
+        else:
+            # direct mapping: e.g. "SHO" → R=S, G=H, B=O
+            letters = list(mode)
+            if len(letters) != 3 or any(l not in ("S","H","O") for l in letters):
+                # invalid code → fallback to natural
+                return self._build_color_composite.__wrapped__(self)
+
+            R = getf(letters[0])
+            G = getf(letters[1])
+            B = getf(letters[2])
+
+        return np.stack([R, G, B], axis=2)
+
+
+    def select_folder(self):
+        folder = QFileDialog.getExistingDirectory(self, "Select Folder to Watch")
+        if folder:
+            self.watch_folder = folder
+            self.folder_label.setText(f"Folder: {os.path.basename(folder)}")
+
+    def load_masters(self):
+        """
+        When the user picks “Load Master Dark…” or “Load Master Flat…”, we load exactly one file
+        (the first in the dialog).  We simply store it in `self.master_dark` or `self.master_flat`,
+        but we also check its dimensions against any existing master so that the user can’t load
+        a 2D flat while the dark is 3D (for example).
+        """
+        sender = self.sender()
+        dlg = QFileDialog(self, "Select Master Files",
+                         filter="FITS TIFF or XISF (*.fit *.fits *.tif *.tiff *.xisf)")
+        dlg.setFileMode(QFileDialog.FileMode.ExistingFiles)
+        if not dlg.exec():
+            return
+
+        chosen = dlg.selectedFiles()[0]
+        img, hdr, bit_depth, is_mono = load_image(chosen)
+        if img is None:
+            QMessageBox.warning(self, "Load Error",
+                                f"Failed to load master file:\n{chosen}")
+            return
+
+        # Convert everything to float32 for consistency
+        img = img.astype(np.float32)
+
+        if "Dark" in sender.text():
+            # If a flat is already loaded, ensure shape‐compatibility
+            if self.master_flat is not None:
+                if not self._shapes_compatible(master=img, other=self.master_flat):
+                    QMessageBox.warning(
+                        self, "Shape Mismatch",
+                        "Cannot load this master dark: it has incompatible shape "
+                        "vs. the already‐loaded master flat."
+                    )
+                    return
+
+            self.master_dark = img
+            self.dark_status_label.setText("Dark: ✅")
+            self.dark_status_label.setStyleSheet("color: #00cc66; font-weight: bold;")            
+            QMessageBox.information(
+                self, "Master Dark Loaded",
+                f"Loaded master dark:\n{os.path.basename(chosen)}"
+            )
+        else:
+            # "Flat" was clicked
+            if self.master_dark is not None:
+                if not self._shapes_compatible(master=self.master_dark, other=img):
+                    QMessageBox.warning(
+                        self, "Shape Mismatch",
+                        "Cannot load this master flat: it has incompatible shape "
+                        "vs. the already‐loaded master dark."
+                    )
+                    return
+
+            self.master_flat = img
+            self.flat_status_label.setText("Flat: ✅")
+            self.flat_status_label.setStyleSheet("color: #00cc66; font-weight: bold;")            
+            QMessageBox.information(
+                self, "Master Flat Loaded",
+                f"Loaded master flat:\n{os.path.basename(chosen)}"
+            )
+
+    def load_filter_flats(self):
+        """
+        Let the user pick one or more flat files.
+        We try to read the FITS header FILTER key to decide which filter
+        each flat belongs to; otherwise fall back to the filename.
+        """
+        dlg = QFileDialog(self, "Select Filter Flats",
+                          filter="FITS or TIFF (*.fit *.fits *.tif *.tiff)")
+        dlg.setFileMode(QFileDialog.FileMode.ExistingFiles)
+        if not dlg.exec():
+            return
+
+        files = dlg.selectedFiles()
+        loaded = []
+        for path in files:
+            img, hdr, bit_depth, is_mono = load_image(path)
+            if img is None:
+                continue
+            # guess filter key from header, else from filename
+            key = None
+            if hdr and hdr.get("FILTER"):
+                key = self._get_filter_key(hdr)
+            if not key:
+                # fallback: basename before extension
+                key = os.path.splitext(os.path.basename(path))[0]
+
+            # store it
+            self.master_flats[key] = img.astype(np.float32)
+            loaded.append(key)
+
+        # update the flat status label to list loaded filters
+        if loaded:
+            names = ", ".join(loaded)
+            self.flat_status_label.setText(f"Flats: {names}")
+            self.flat_status_label.setStyleSheet("color: #00cc66; font-weight: bold;")
+            QMessageBox.information(
+                self, "Filter Flats Loaded",
+                f"Loaded flats for filters: {names}"
+            )
+        else:
+            QMessageBox.warning(self, "No Flats Loaded",
+                                "No flats could be loaded.")
+
+    def _shapes_compatible(self, master: np.ndarray, other: np.ndarray) -> bool:
+        """
+        Return True if `master` and `other` can be used together in calibration:
+          - Exactly the same shape, OR
+          - master is 2D (H×W) and other is 3D (H×W×3), OR
+          - vice versa.
+        """
+        if master.shape == other.shape:
+            return True
+
+        # If one is 2D and the other is H×W×3, check the first two dims
+        if master.ndim == 2 and other.ndim == 3 and other.shape[:2] == master.shape:
+            return True
+        if other.ndim == 2 and master.ndim == 3 and master.shape[:2] == other.shape:
+            return True
+
+        return False
+
+    def _average_images(self, paths):
+        # stub: load each via load_image(), convert to float32, accumulate & divide
+        return None
+
+    def _normalized_average(self, paths):
+        # stub: load each, divide by its mean, average them, then renormalize
+        return None
+
+    def start_and_process(self):
+        """Process everything currently in folder, then begin monitoring."""
+        if not self.watch_folder:
+            self.status_label.setText("❗ No folder selected")
+            return
+        # Clear any old record so existing files are re-processed
+        self.processed_files.clear()
+        # Process all current files once
+        self.check_for_new_frames()
+        # Now start monitoring
+        self.is_running = True
+        self.poll_timer.start()
+        self.status_label.setText(f"▶ Processing & Monitoring: {os.path.basename(self.watch_folder)}")
+
+    def start_monitor_only(self):
+        """Mark existing files as seen and only process new arrivals."""
+        if not self.watch_folder:
+            self.status_label.setText("❗ No folder selected")
+            return
+        # Populate processed_files with all existing files so they won't be re-processed
+        exts = (
+            "*.fit", "*.fits", "*.tif", "*.tiff",
+            "*.cr2", "*.cr3", "*.nef", "*.arw",
+            "*.dng", "*.raf", "*.orf", "*.rw2", "*.pef", "*.xisf", "*.png", "*.jpg", "*.jpeg"
+        )
+        all_paths = []
+        for ext in exts:
+            all_paths += glob.glob(os.path.join(self.watch_folder, "**", ext), recursive=True)
+        self.processed_files = set(all_paths)
+
+        # Start monitoring
+        self.is_running = True
+        self.poll_timer.start()
+        self.status_label.setText(f"▶ Monitoring Only: {os.path.basename(self.watch_folder)}")
+
+    def start_live(self):
+        if not self.watch_folder:
+            self.status_label.setText("❗ No folder selected")
+            return
+        self.is_running = True
+        self.poll_timer.start()
+        self.status_label.setText(f"▶ Monitoring: {os.path.basename(self.watch_folder)}")
+        self.mode_label.setText("Mode: Linear Average")
+
+    def stop_live(self):
+        if self.is_running:
+            self.is_running = False
+            self.poll_timer.stop()
+            self.status_label.setText("■ Stopped")
+        else:
+            self.status_label.setText("■ Already stopped")
+
+    def reset_live(self):
+        if self.is_running:
+            self.is_running = False
+            self.poll_timer.stop()
+            self.status_label.setText("■ Stopped")
+        else:
+            self.status_label.setText("■ Already stopped")
+
+        # Clear all state
+        self.processed_files.clear()
+        self.frame_count = 0
+        self.current_stack = None
+
+        self.total_exposure = 0.0
+        self.exposure_label.setText("Total Exp: 00:00:00")
+
+        self.filter_stacks.clear()
+        self.filter_counts.clear()
+        self.filter_buffers.clear()
+        self.filter_mus.clear()
+        self.filter_m2s.clear()
+
+        if hasattr(self, 'reference_image_2d'):
+            del self.reference_image_2d
+
+        # Re-initialize bootstrapping stats
+        self._buffer = []
+        self._mu = None
+        self._m2 = None
+
+        # NEW: clear the metrics panel
+        self.metrics_window.metrics_panel.clear_all()
+
+        # Update labels
+        self.frame_count_label.setText("Frames: 0")
+        self.status_label.setText("↺ Reset")
+        self.mode_label.setText("Mode: Linear Average")
+
+        # Clear the displayed image
+        self.pixmap_item.setPixmap(QPixmap())
+
+        # Reset zoom/pan fit flag
+        self._did_initial_fit = False
+        #self.master_dark = None
+        #self.master_flat = None
+        #self.dark_status_label.setText("Dark: ❌")
+        #self.flat_status_label.setText("Flat: ❌")
+        #self.dark_status_label.setStyleSheet("color: #cccccc; font-weight: bold;")
+        #self.flat_status_label.setStyleSheet("color: #cccccc; font-weight: bold;")        
+
+
+    def _update_probe(self, path: str) -> dict:
+        """Update probe info (size, mtime) for path and return the info dict."""
+        try:
+            st = os.stat(path)
+        except FileNotFoundError:
+            # file disappeared; clear any probe info
+            self._probe.pop(path, None)
+            return None
+        now = time.time()
+        size, mtime = st.st_size, st.st_mtime
+
+        info = self._probe.get(path)
+        if info is None:
+            info = {"size": size, "mtime": mtime, "since": now, "penalty_until": 0.0}
+            self._probe[path] = info
+            return info
+
+        # If size or mtime changed, reset stability timer
+        if size != info["size"] or mtime != info["mtime"]:
+            info["size"] = size
+            info["mtime"] = mtime
+            info["since"] = now
+        return info
+
+    def _can_open_for_read(self, path: str) -> bool:
+        """
+        Try a tiny open+read to ensure the writer has released the handle.
+        If we hit PermissionError / OSError, we mark a penalty and say 'not ready'.
+        """
+        try:
+            with open(path, "rb") as f:
+                _ = f.read(1)
+            return True
+        except (PermissionError, OSError):
+            # mark a penalty window so we don't hammer the file immediately
+            info = self._probe.get(path) or self._update_probe(path)
+            if info:
+                info["penalty_until"] = time.time() + self.OPEN_RETRY_PENALTY_SECS
+            return False
+
+    def _file_ready(self, path: str) -> bool:
+        """
+        A file is 'ready' when:
+          - we are not inside a penalty window,
+          - size+mtime have been unchanged for FILE_STABLE_SECS,
+          - and we can actually open it for reading.
+        """
+        info = self._update_probe(path)
+        if info is None:
+            return False  # missing
+
+        now = time.time()
+        if now < info.get("penalty_until", 0.0):
+            return False
+
+        # Require size+mtime to be unchanged for FILE_STABLE_SECS
+        if (now - info["since"]) < self.FILE_STABLE_SECS:
+            return False
+
+        # Finally confirm we can open the file (this also sets penalty if it fails)
+        return self._can_open_for_read(path)
+
+
+    def check_for_new_frames(self):
+        if not self.is_running or not self.watch_folder:
+            return
+
+        # Gather candidates
+        exts = (
+            "*.fit", "*.fits", "*.tif", "*.tiff",
+            "*.cr2", "*.cr3", "*.nef", "*.arw",
+            "*.dng", "*.raf", "*.orf", "*.rw2", "*.pef", "*.xisf",
+            "*.png", "*.jpg", "*.jpeg"
+        )
+        all_paths = []
+        for ext in exts:
+            all_paths += glob.glob(os.path.join(self.watch_folder, '**', ext), recursive=True)
+
+        # Only consider paths not yet processed
+        candidates = [p for p in sorted(all_paths) if p not in self.processed_files]
+        if not candidates:
+            return
+
+        # Show first new file name (status only)
+        self.status_label.setText(f"➜ New/updated files: {len(candidates)}")
+        QApplication.processEvents()
+
+        # Probe each candidate: only process when 'ready'
+        processed_now = 0
+        for path in candidates:
+            # Skip if we recently penalized this path
+            info = self._probe.get(path)
+            if info and time.time() < info.get("penalty_until", 0.0):
+                continue
+
+            # Check readiness: stable size/mtime and can open-for-read
+            if not self._file_ready(path):
+                continue  # not yet ready; we'll see it again on the next tick
+
+            # Only *now* do we mark as processed and actually process the frame
+            self.processed_files.add(path)
+            base = os.path.basename(path)
+            self.status_label.setText(f"→ Processing: {base}")
+            QApplication.processEvents()
+
+            try:
+                self.process_frame(path)
+                processed_now += 1
+            except Exception as e:
+                # If anything unexpected happens, clear 'processed' so we can retry later
+                # but add a penalty to avoid tight loops.
+                self.processed_files.discard(path)
+                info = self._probe.get(path) or self._update_probe(path)
+                if info:
+                    info["penalty_until"] = time.time() + self.OPEN_RETRY_PENALTY_SECS
+                self.status_label.setText(f"⚠ Error on {base}: {e}")
+                QApplication.processEvents()
+
+        if processed_now > 0:
+            self.status_label.setText(f"✔ Processed {processed_now} file(s)")
+            QApplication.processEvents()
+
+    def process_frame(self, path):
+        if not self._file_ready(path):
+            # do not mark as processed here; monitor will retry after cool-down
+            return
+
+        # if star-trail mode is on, bypass the normal pipeline entirely:
+        if self.star_trail_mode:
+            return self._process_star_trail(path)
+                
+        # 1) Load
+        # ─── 1) RAW‐file check ────────────────────────────────────────────
+        lower = path.lower()
+        raw_exts = ('.cr2', '.cr3', '.nef', '.arw', '.dng', '.raf', '.orf', '.rw2', '.pef')
+        if lower.endswith(raw_exts):
+            # Attempt to decode using rawpy
+            try:
+                with rawpy.imread(path) as raw:
+                    # Postprocess into an 8‐bit RGB array
+                    # (you could tweak postprocess params if desired)
+                    img_rgb8 = raw.postprocess(
+                        use_camera_wb=True,
+                        no_auto_bright=True,
+                        output_bps=16
+                    )  # shape (H, W, 3), dtype=uint8
+
+                # Convert to float32 [0..1] so it matches load_image() behavior
+                img = img_rgb8.astype(np.float32) / 65535.0
+
+                # Build a minimal FITS header and attempt to extract EXIF tags
+                header = fits.Header()
+                header["SIMPLE"] = True
+                header["BITPIX"] = 16
+                header["CREATOR"] = "LiveStack(RAW)"
+                header["IMAGETYP"] = "RAW"
+                # Default EXPTIME/ISO/DATE-OBS in case EXIF fails
+                header["EXPTIME"] = "Unknown"
+                header["ISO"]     = "Unknown"
+                header["DATE-OBS"] = "Unknown"
+
+                try:
+                    with open(path, 'rb') as f:
+                        tags = exifread.process_file(f, details=False)
+                    # EXIF: ExposureTime
+                    exp_tag = tags.get("EXIF ExposureTime") or tags.get("EXIF ShutterSpeedValue")
+                    if exp_tag:
+                        exp_str = str(exp_tag.values)
+                        if '/' in exp_str:
+                            top, bot = exp_str.split('/', 1)
+                            header["EXPTIME"] = (float(top)/float(bot), "Exposure Time (s)")
+                        else:
+                            header["EXPTIME"] = (float(exp_str), "Exposure Time (s)")
+                    # ISO
+                    iso_tag = tags.get("EXIF ISOSpeedRatings")
+                    if iso_tag:
+                        header["ISO"] = str(iso_tag.values)
+                    # Date/time original
+                    date_obs = tags.get("EXIF DateTimeOriginal")
+                    if date_obs:
+                        header["DATE-OBS"] = str(date_obs.values)
+                except Exception:
+                    # If EXIF parsing fails, just leave defaults
+                    pass
+
+                bit_depth = 16
+                is_mono = False
+
+            except Exception as e:
+                # If rawpy fails, bail out early
+                self.status_label.setText(f"⚠ Failed to decode RAW: {os.path.basename(path)}")
+                QApplication.processEvents()
+                return
+
+        else:
+            # ─── 2) Not RAW → call your existing load_image()
+            img, header, bit_depth, is_mono = load_image(path)
+            if img is None:
+                self.status_label.setText(f"⚠ Failed to load {os.path.basename(path)}")
+                QApplication.processEvents()
+                return
+
+        # ——— 2) CALIBRATION (once) ————————————————————————
+        # ——— 2a) DETECT MONO→COLOR MODE ————————————————————
+        mono_key = None
+        if self.mono_color_mode and is_mono and header.get('FILTER') and 'BAYERPAT' not in header:
+            mono_key = self._get_filter_key(header)
+
+        # ——— 2b) CALIBRATION (once) ————————————————————————
+        if self.master_dark is not None:
+            img = img.astype(np.float32) - self.master_dark
+        # prefer per-filter flat if we’re in mono→color and have one
+        if mono_key and mono_key in self.master_flats:
+            img = apply_flat_division_numba(img, self.master_flats[mono_key])
+        elif self.master_flat is not None:
+            img = apply_flat_division_numba(img, self.master_flat)
+
+        # ——— 3) DEBAYER if BAYERPAT ——————————————————————
+        if is_mono and header.get('BAYERPAT'):
+            pat = header['BAYERPAT'][0] if isinstance(header['BAYERPAT'], tuple) else header['BAYERPAT']
+            img = debayer_fits_fast(img, pat)
+            is_mono = False
+
+        # ——— 5) PROMOTION TO 3-CHANNEL if NOT in mono-mode —————
+        if mono_key is None and img.ndim == 2:
+            img = np.stack([img, img, img], axis=2)
+
+        # ——— 6) BUILD PLANE for alignment & metrics —————————
+        plane = img if (mono_key and img.ndim == 2) else np.mean(img, axis=2)
+
+        # ——— 7) ALIGN to reference_image_2d ——————————————————
+        if hasattr(self, 'reference_image_2d'):
+            delta = StarRegistrationWorker.compute_affine_transform_astroalign(
+                plane, self.reference_image_2d
+            )
+            if delta is None:
+                delta = IDENTITY_2x3
+            # apply to full img (if color) and to plane
+            if mono_key is None:
+                img = StarRegistrationThread.apply_affine_transform_static(img, delta)
+            plane = StarRegistrationThread.apply_affine_transform_static(
+                plane if plane.ndim == 2 else plane[:, :, None], delta
+            ).squeeze()
+
+        # ——— 8) NORMALIZE —————————————————————————————
+        if mono_key:
+            norm_plane = stretch_mono_image(plane, target_median=0.3)
+            norm_color = None
+        else:
+            norm_color = stretch_color_image(img, target_median=0.3, linked=False)
+            norm_plane = np.mean(norm_color, axis=2)
+
+        # ——— 9) METRICS & SNR —————————————————————————
+        sc, fwhm, ecc = compute_frame_star_metrics(norm_plane)
+        # instead, use the cumulative stack (or composite) for SNR:
+        if mono_key:
+            # once we have any filter_stacks, build the composite;
+            # fall back to this frame’s plane if it’s the first one
+            if self.filter_stacks:
+                stack_img = self._build_color_composite()
+            else:
+                stack_img = norm_plane
+        else:
+            # for color‐only, use the running‐average stack once it exists,
+            # else fall back to this frame’s normalized color
+            if self.current_stack is not None:
+                stack_img = self.current_stack
+            else:
+                stack_img = norm_color
+        snr_val = estimate_global_snr(stack_img)
+
+        # ——— 10) CULLING? ————————————————————————————
+        flagged = (
+            (fwhm > self.max_fwhm) or
+            (ecc > self.max_ecc)     or
+            (sc < self.min_star_count)
+        )
+        if flagged:
+            self._cull_frame(path)
+            self.metrics_window.metrics_panel.add_point(
+                self.frame_count + 1, fwhm, ecc, sc, snr_val, True
+            )
+            return
+
+        # ─── 11) FIRST-FRAME INITIALIZATION ──────────────────────────────
+        if self.frame_count == 0:
+            # set reference on the very first good frame
+            self.reference_image_2d = norm_plane.copy()
+            self.frame_count = 1
+            self.frame_count_label.setText("Frames: 1")
+            # always start in linear‐average mode
+            if mono_key:
+                self.mode_label.setText(f"Mode: Linear Average ({mono_key})")
+                self.status_label.setText(f"Started {mono_key}-filter linear stack")
+            else:
+                self.mode_label.setText("Mode: Linear Average")
+                self.status_label.setText("Started linear stack")
+            QApplication.processEvents()
+
+            if mono_key:
+                # start the filter stack
+                self.filter_stacks[mono_key]  = norm_plane.copy()
+                self.filter_counts[mono_key]  = 1
+                self.filter_buffers[mono_key] = [norm_plane.copy()]
+            else:
+                # start the normal running stack
+                self.current_stack = norm_color.copy()
+                self._buffer = [norm_color.copy()]
+            # ─── accumulate exposure ─────────────────────
+            exp_val = header.get("EXPOSURE", header.get("EXPTIME", None))
+            if exp_val is not None:
+                try:
+                    secs = float(exp_val)
+                    self.total_exposure += secs
+                    hrs  = int(self.total_exposure // 3600)
+                    mins = int((self.total_exposure % 3600) // 60)
+                    secs_rem = int(self.total_exposure % 60)
+                    self.exposure_label.setText(
+                        f"Total Exp: {hrs:02d}:{mins:02d}:{secs_rem:02d}"
+                    )
+                except Exception:
+                    pass  # Ignore exposure parsing errors
+            QApplication.processEvents()
+
+
+        else:
+            # ─── 12) RUNNING–AVERAGE or CLIP-σ UPDATE ────────────────────
+            if mono_key is None:
+                # — Color-only stacking —
+                if self.frame_count < self.bootstrap_frames:
+                    # 12a) Linear bootstrap
+                    n = self.frame_count + 1
+                    self.current_stack = (
+                        (self.frame_count / n) * self.current_stack
+                        + (1.0 / n) * norm_color
+                    )
+                    self._buffer.append(norm_color.copy())
+
+                    # hit the bootstrap threshold?
+                    if n == self.bootstrap_frames:
+                        # init Welford stats
+                        buf = np.stack(self._buffer, axis=0)
+                        self._mu = np.mean(buf, axis=0)
+                        diffs = buf - self._mu[np.newaxis, ...]
+                        self._m2 = np.sum(diffs * diffs, axis=0)
+                        self._buffer = None
+
+                        # switch to clipping mode
+                        self.mode_label.setText("Mode: μ-σ Clipping Average")
+                        self.status_label.setText("Switched to μ–σ clipping (color)")
+                        QApplication.processEvents()
+                    else:
+                        # still linear
+                        self.mode_label.setText("Mode: Linear Average")
+                        self.status_label.setText(f"Processed color frame #{n} (linear)")
+                        QApplication.processEvents()
+                else:
+                    # 12b) μ–σ clipping
+                    sigma = np.sqrt(self._m2 / (self.frame_count - 1))
+                    mask = np.abs(norm_color - self._mu) <= (self.clip_threshold * sigma)
+                    clipped = np.where(mask, norm_color, self._mu)
+
+                    n = self.frame_count + 1
+                    self.current_stack = (
+                        (self.frame_count / n) * self.current_stack
+                        + (1.0 / n) * clipped
+                    )
+
+                    # Welford update
+                    delta_mu = clipped - self._mu
+                    self._mu += delta_mu / n
+                    delta2 = clipped - self._mu
+                    self._m2 += delta_mu * delta2
+
+                    # stay in clipping mode
+                    self.mode_label.setText("Mode: μ-σ Clipping Average")
+                    self.status_label.setText(f"Processed color frame #{n} (clipped)")
+                    QApplication.processEvents()
+
+                # bump global frame count
+                self.frame_count = n
+                # ─── accumulate exposure ─────────────────────
+                exp_val = header.get("EXPOSURE", header.get("EXPTIME", None))
+                if exp_val is not None:
+                    try:
+                        secs = float(exp_val)
+                        self.total_exposure += secs
+                        hrs  = int(self.total_exposure // 3600)
+                        mins = int((self.total_exposure % 3600) // 60)
+                        secs_rem = int(self.total_exposure % 60)
+                        self.exposure_label.setText(
+                            f"Total Exp: {hrs:02d}:{mins:02d}:{secs_rem:02d}"
+                        )
+                    except Exception:
+                        pass  # Ignore exposure parsing errors
+                QApplication.processEvents()
+
+
+            else:
+                # — Mono→color (per-filter) stacking —
+                count = self.filter_counts.get(mono_key, 0)
+                buf   = self.filter_buffers.setdefault(mono_key, [])
+
+                if count < self.bootstrap_frames:
+                    # 12c) Linear bootstrap per-filter
+                    new_count = count + 1
+                    if count == 0:
+                        self.filter_stacks[mono_key] = norm_plane.copy()
+                    else:
+                        self.filter_stacks[mono_key] = (
+                            (count / new_count) * self.filter_stacks[mono_key]
+                            + (1.0 / new_count) * norm_plane
+                        )
+                    buf.append(norm_plane.copy())
+                    self.filter_counts[mono_key] = new_count
+
+                    if new_count == self.bootstrap_frames:
+                        # init Welford
+                        stacked = np.stack(buf, axis=0)
+                        mu    = np.mean(stacked, axis=0)
+                        diffs = stacked - mu[np.newaxis, ...]
+                        m2    = np.sum(diffs * diffs, axis=0)
+                        self.filter_mus[mono_key] = mu
+                        self.filter_m2s[mono_key] = m2
+
+                        self.mode_label.setText(f"Mode: μ-σ Clipping Average ({mono_key})")
+                        self.status_label.setText(f"Switched to μ–σ clipping ({mono_key})")
+                        QApplication.processEvents()
+                    else:
+                        # still linear
+                        self.mode_label.setText(f"Mode: Linear Average ({mono_key})")
+                        self.status_label.setText(
+                            f"Processed {mono_key}-filter frame #{new_count} (linear)"
+                        )
+                        QApplication.processEvents()
+
+                else:
+                    # 12d) μ–σ clipping per-filter
+                    mu = self.filter_mus[mono_key]
+                    m2 = self.filter_m2s[mono_key]
+                    sigma = np.sqrt(m2 / (count - 1))
+                    mask   = np.abs(norm_plane - mu) <= (self.clip_threshold * sigma)
+                    clipped = np.where(mask, norm_plane, mu)
+
+                    new_count = count + 1
+                    self.filter_stacks[mono_key] = (
+                        (count / new_count) * self.filter_stacks[mono_key]
+                        + (1.0 / new_count) * clipped
+                    )
+
+                    # Welford update on µ and m2
+                    delta   = clipped - mu
+                    new_mu  = mu + delta / new_count
+                    delta2  = clipped - new_mu
+                    new_m2  = m2 + delta * delta2
+                    self.filter_mus[mono_key] = new_mu
+                    self.filter_m2s[mono_key] = new_m2
+                    self.filter_counts[mono_key] = new_count
+
+                    self.mode_label.setText(f"Mode: μ-σ Clipping Average ({mono_key})")
+                    self.status_label.setText(
+                        f"Processed {mono_key}-filter frame #{new_count} (clipped)"
+                    )
+                    QApplication.processEvents()
+
+                # bump global frame count
+                self.frame_count += 1
+                self.frame_count_label.setText(f"Frames: {self.frame_count}")
+                # ─── accumulate exposure ─────────────────────
+                exp_val = header.get("EXPOSURE", header.get("EXPTIME", None))
+                if exp_val is not None:
+                    try:
+                        secs = float(exp_val)
+                        self.total_exposure += secs
+                        hrs  = int(self.total_exposure // 3600)
+                        mins = int((self.total_exposure % 3600) // 60)
+                        secs_rem = int(self.total_exposure % 60)
+                        self.exposure_label.setText(
+                            f"Total Exp: {hrs:02d}:{mins:02d}:{secs_rem:02d}"
+                        )
+                    except Exception:
+                        pass  # Ignore exposure parsing errors
+                QApplication.processEvents()
+
+            # ─── 13) Update UI ─────────────────────────────────────────
+            self.frame_count_label.setText(f"Frames: {self.frame_count}")
+            QApplication.processEvents()
+
+        # ——— 13) METRICS PANEL for good frame —————————————
+        self.metrics_window.metrics_panel.add_point(
+            self.frame_count, fwhm, ecc, sc, snr_val, False
+        )
+
+        # ——— 14) PREVIEW & STATUS LABEL —————————————————————
+        if mono_key:
+            preview = self._build_color_composite()
+            self.status_label.setText(f"Stacked {mono_key}-filter frame {os.path.basename(path)}")
+            QApplication.processEvents()
+        else:
+            preview = self.current_stack
+            self.status_label.setText(f"✔ processed {os.path.basename(path)}")
+            QApplication.processEvents()
+
+        self.update_preview(preview)
+        QApplication.processEvents()
+
+    def _process_star_trail(self, path: str):
+        """
+        Load/calibrate a single frame (RAW or FITS/TIFF), debayer if needed,
+        normalize, then build a max‐value “star trail” in self.current_stack.
+        """
+        # ─── 1) Load (RAW vs FITS) ─────────────────────────────
+        lower = path.lower()
+        raw_exts = ('.cr2', '.cr3', '.nef', '.arw', '.dng', '.raf',
+                    '.orf', '.rw2', '.pef')
+        if lower.endswith(raw_exts):
+            try:
+                with rawpy.imread(path) as raw:
+                    img_rgb8 = raw.postprocess(use_camera_wb=True,
+                                               no_auto_bright=True,
+                                               output_bps=16)
+                img = img_rgb8.astype(np.float32) / 65535.0
+                header = fits.Header()
+                header["SIMPLE"]   = True
+                header["BITPIX"]   = 16
+                header["CREATOR"]  = "LiveStack(RAW)"
+                header["IMAGETYP"] = "RAW"
+                header["EXPTIME"]  = "Unknown"
+                # attempt EXIF, same as process_frame…
+                try:
+                    with open(path,'rb') as f:
+                        tags = exifread.process_file(f, details=False)
+                    exp_tag = tags.get("EXIF ExposureTime") \
+                              or tags.get("EXIF ShutterSpeedValue")
+                    if exp_tag:
+                        ev = str(exp_tag.values)
+                        if '/' in ev:
+                            n,d = ev.split('/',1)
+                            header["EXPTIME"] = (float(n)/float(d),
+                                                 "Exposure Time (s)")
+                        else:
+                            header["EXPTIME"] = (float(ev),
+                                                 "Exposure Time (s)")
+                except Exception:
+                    pass  # Ignore EXIF parsing errors
+                bit_depth = 16
+                is_mono   = False
+            except Exception:
+                self.status_label.setText(
+                    f"⚠ Failed to decode RAW: {os.path.basename(path)}"
+                )
+                QApplication.processEvents()
+                return
+        else:
+            # FITS / TIFF / XISF
+            img, header, bit_depth, is_mono = load_image(path)
+            if img is None:
+                self.status_label.setText(
+                    f"⚠ Failed to load {os.path.basename(path)}"
+                )
+                QApplication.processEvents()
+                return
+
+        # ─── 2) Calibration ─────────────────────────────────────
+        mono_key = None
+        if (self.mono_color_mode
+            and is_mono
+            and header.get('FILTER')
+            and 'BAYERPAT' not in header):
+            mono_key = self._get_filter_key(header)
+
+        if self.master_dark is not None:
+            img = img.astype(np.float32) - self.master_dark
+
+        if mono_key and mono_key in self.master_flats:
+            img = apply_flat_division_numba(img,
+                                            self.master_flats[mono_key])
+        elif self.master_flat is not None:
+            img = apply_flat_division_numba(img,
+                                            self.master_flat)
+
+        # ─── 3) Debayer ─────────────────────────────────────────
+        if is_mono and header.get('BAYERPAT'):
+            pat = (header['BAYERPAT'][0]
+                   if isinstance(header['BAYERPAT'], tuple)
+                   else header['BAYERPAT'])
+            img = debayer_fits_fast(img, pat)
+            is_mono = False
+
+        # ─── 4) Force 3-channel if still mono ───────────────────
+        if not mono_key and img.ndim == 2:
+            img = np.stack([img, img, img], axis=2)
+
+        # ─── 5) Normalize ───────────────────────────────────────
+        # for star-trail we want a visible, stretched version:
+        if img.ndim == 2:
+            plane = stretch_mono_image(img, target_median=0.3)
+            norm_color = np.stack([plane]*3, axis=2)
+        else:
+            norm_color = stretch_color_image(img,
+                                             target_median=0.3,
+                                             linked=False)
+
+        # ─── 6) Build max-value stack ───────────────────────────
+        if self.frame_count == 0:
+            self.current_stack = norm_color.copy()
+        else:
+            # elementwise max over all frames so far
+            self.current_stack = np.maximum(self.current_stack,
+                                            norm_color)
+
+        # ─── 7) Update counters and labels ──────────────────────
+        self.frame_count += 1
+        self.frame_count_label.setText(f"Frames: {self.frame_count}")
+
+        exp_val = header.get("EXPOSURE", header.get("EXPTIME", None))
+        if exp_val is not None:
+            try:
+                secs = float(exp_val)
+                self.total_exposure += secs
+                h = int(self.total_exposure // 3600)
+                m = int((self.total_exposure % 3600)//60)
+                s = int(self.total_exposure % 60)
+                self.exposure_label.setText(
+                    f"Total Exp: {h:02d}:{m:02d}:{s:02d}")
+            except Exception:
+                pass  # Ignore exposure parsing errors
+
+        self.status_label.setText(
+            f"★ Star-Trail frame {self.frame_count}: "
+            f"{os.path.basename(path)}"
+        )
+        self.update_preview(self.current_stack)
+        QApplication.processEvents()
+
+
+
+    def update_preview(self, array: np.ndarray):
+        # 1) normalize, apply contrast/brightness
+        arr = np.clip(array, 0.0, 1.0).astype(np.float32)
+        pivot = 0.3
+        arr = ((arr - pivot) * self.contrast + pivot) + self.brightness
+        arr = np.clip(arr, 0.0, 1.0)
+
+        # 2) convert to uint8 and KEEP a reference on self
+        self._last_frame_bytes = (arr * 255).astype(np.uint8)
+        h, w = self._last_frame_bytes.shape[:2]
+
+        # 3) build QImage from the kept buffer
+        if self._last_frame_bytes.ndim == 2:
+            fmt = QImage.Format.Format_Grayscale8
+            bytespp = w
+        else:
+            fmt = QImage.Format.Format_RGB888
+            bytespp = 3 * w
+        qimg = QImage(self._last_frame_bytes.data, w, h, bytespp, fmt)
+
+        # 4) update scene
+        self.pixmap_item.setPixmap(QPixmap.fromImage(qimg))
+        self.scene.setSceneRect(0, 0, w, h)
+
+        if not self._did_initial_fit:
+            self.view.fitInView(self.scene.sceneRect(), Qt.AspectRatioMode.KeepAspectRatio)
+            self._did_initial_fit = True
+