setiastrosuitepro 1.6.7__py3-none-any.whl

setiastro/saspro/perfect_palette_picker.py

@@ -0,0 +1,1105 @@
+# pro/perfect_palette_picker.py
+from __future__ import annotations
+import os
+import numpy as np
+from PIL import Image
+import cv2
+from PyQt6.QtCore import Qt, QSize, QEvent, QTimer, QPoint, pyqtSignal
+from PyQt6.QtWidgets import (
+    QWidget, QVBoxLayout, QHBoxLayout, QLabel, QPushButton, QScrollArea,
+    QFileDialog, QInputDialog, QMessageBox, QGridLayout, QCheckBox, QSizePolicy, QDialog
+)
+from PyQt6.QtGui import QPixmap, QImage, QIcon, QPainter, QPen, QColor, QFont, QFontMetrics, QCursor
+
+# legacy loader (same one DocManager uses)
+from setiastro.saspro.legacy.image_manager import load_image as legacy_load_image
+
+# your statistical stretch (mono + color) like SASv2
+# (same signatures you use elsewhere)
+from setiastro.saspro.imageops.stretch import stretch_mono_image, stretch_color_image
+from setiastro.saspro.widgets.themed_buttons import themed_toolbtn
+
+class PaletteAdjustDialog(QDialog):
+    adjusted_image = pyqtSignal(np.ndarray)
+
+    def __init__(self, base_rgb, palette_name, ha_src, oiii_src, sii_src, owner):
+        from PyQt6.QtWidgets import QDialog, QVBoxLayout, QHBoxLayout, QLabel, QPushButton, QScrollArea, QSlider
+        from PyQt6.QtCore import QTimer, Qt, QPoint, QEvent
+        super().__init__(owner)
+        self.setWindowTitle("Adjust Palette Intensities")
+        self.setWindowFlag(Qt.WindowType.Window, True)
+        self.setWindowModality(Qt.WindowModality.NonModal)
+        self.setModal(False)
+        #self.setAttribute(Qt.WidgetAttribute.WA_DeleteOnClose, True)
+
+        self.base_rgb     = base_rgb.astype(np.float32)
+        self.palette_name = palette_name
+        self.ha_src       = ha_src
+        self.oiii_src     = oiii_src
+        self.sii_src      = sii_src
+        self.owner        = owner
+
+        self.ha_factor = 1.0
+        self.oiii_factor = 1.0
+        self.sii_factor = 1.0
+
+        self._debounce = QTimer(self); self._debounce.setInterval(300); self._debounce.setSingleShot(True)
+        self._debounce.timeout.connect(self._update_preview)
+
+        self.zoom_factor = 1.0
+        self._dragging = False
+        self._last_pos = QPoint()
+
+        vlayout = QVBoxLayout(self)
+
+        # Zoom controls
+        zoom_layout = QHBoxLayout()
+
+        self.btn_zoom_in  = themed_toolbtn("zoom-in", "Zoom In")
+        self.btn_zoom_out = themed_toolbtn("zoom-out", "Zoom Out")
+        self.btn_fit      = themed_toolbtn("zoom-fit-best", "Fit to Preview")
+
+        self.btn_zoom_in.clicked.connect(lambda: self._change_zoom(1.25))
+        self.btn_zoom_out.clicked.connect(lambda: self._change_zoom(0.8))
+        self.btn_fit.clicked.connect(self._fit_to_preview)
+
+        zoom_layout.addStretch(1)
+        zoom_layout.addWidget(self.btn_zoom_out)
+        zoom_layout.addWidget(self.btn_zoom_in)
+        zoom_layout.addWidget(self.btn_fit)
+        zoom_layout.addStretch(1)
+
+        vlayout.addLayout(zoom_layout)
+
+        # Preview
+        self.preview_area = QScrollArea(self); self.preview_area.setWidgetResizable(True)
+        self.preview_label = QLabel(alignment=Qt.AlignmentFlag.AlignCenter)
+        self.preview_label.setCursor(Qt.CursorShape.OpenHandCursor)
+        self.preview_label.setMouseTracking(True)
+        self.preview_area.setWidget(self.preview_label)
+        self.preview_label.installEventFilter(self)
+        vlayout.addWidget(self.preview_area, stretch=1)
+
+        # Sliders
+        for name in ("Ha","OIII","SII"):
+            row = QHBoxLayout()
+            row.addWidget(QLabel(f"{name} Intensity:", self))
+            sl = QSlider(Qt.Orientation.Horizontal, self); sl.setRange(0,200); sl.setValue(100)
+            sl.valueChanged.connect(self._on_slider_change)
+            setattr(self, f"_{name.lower()}_slider", sl)
+            row.addWidget(sl)
+            vlayout.addLayout(row)
+
+        # Buttons
+        btns = QHBoxLayout(); btns.addStretch()
+        accept  = QPushButton("Accept", self); accept.clicked.connect(self._on_accept)
+        reset   = QPushButton("Reset",  self); reset.clicked.connect(self._on_reset)
+        discard = QPushButton("Discard",self); discard.clicked.connect(self.reject)
+        btns.addWidget(accept); btns.addWidget(reset); btns.addWidget(discard)
+        vlayout.addLayout(btns)
+
+        self._update_preview()
+
+    def _on_slider_change(self, _):
+        self.ha_factor   = self._ha_slider.value()/100.0
+        self.oiii_factor = self._oiii_slider.value()/100.0
+        self.sii_factor  = self._sii_slider.value()/100.0
+        self._debounce.start()
+
+    def _update_preview(self):
+        ha = (self.ha_src   * self.ha_factor)   if self.ha_src   is not None else None
+        oo = (self.oiii_src * self.oiii_factor) if self.oiii_src is not None else None
+        si = (self.sii_src  * self.sii_factor)  if self.sii_src  is not None else None
+
+        r,g,b = self.owner._map_channels_or_special(self.palette_name, ha, oo, si)
+
+        # --- make sure channels match the base palette size ---
+        H, W = self.base_rgb.shape[:2]
+        def fit(ch):
+            if ch is None: return None
+            if ch.shape[:2] != (H, W):
+                return self.owner._resize_to(ch, (W, H))
+            return ch
+        r, g, b = fit(r), fit(g), fit(b)
+        # ------------------------------------------------------
+
+        img = np.zeros_like(self.base_rgb, dtype=np.float32)
+        if r is not None: img[...,0] = r
+        if g is not None: img[...,1] = g
+        if b is not None: img[...,2] = b
+        m = float(img.max()) or 1.0
+        img = np.clip(img/m, 0.0, 1.0)
+
+        qimg = self.owner._to_qimage(img)
+        self._base_pixmap = QPixmap.fromImage(qimg)
+        self._rescale_pixmap()
+
+    def _rescale_pixmap(self):
+        if not hasattr(self, "_base_pixmap"): return
+        w = int(self._base_pixmap.width()  * self.zoom_factor)
+        h = int(self._base_pixmap.height() * self.zoom_factor)
+        scaled = self._base_pixmap.scaled(w, h, Qt.AspectRatioMode.KeepAspectRatio, Qt.TransformationMode.SmoothTransformation)
+        self._current_pixmap = scaled
+        self.preview_label.setPixmap(scaled)
+        self.preview_label.resize(scaled.size())
+
+    def _change_zoom(self, factor: float):
+        self.zoom_factor = max(0.1, min(10.0, self.zoom_factor * factor))
+        self._rescale_pixmap()
+
+    def _fit_to_preview(self):
+        if not hasattr(self, "_base_pixmap"):
+            return
+        vp = self.preview_area.viewport().size()
+        pm = self._base_pixmap.size()
+        if pm.width() <= 0 or pm.height() <= 0:
+            return
+        k = min(vp.width() / pm.width(), vp.height() / pm.height())
+        self.zoom_factor = max(0.1, min(10.0, k))
+        self._rescale_pixmap()
+
+
+    def _on_reset(self):
+        for s in (self._ha_slider, self._oiii_slider, self._sii_slider):
+            s.setValue(100)
+        self._on_slider_change(None)
+
+    def _on_accept(self):
+        ha = (self.ha_src   * self.ha_factor)   if self.ha_src   is not None else None
+        oo = (self.oiii_src * self.oiii_factor) if self.oiii_src is not None else None
+        si = (self.sii_src  * self.sii_factor)  if self.sii_src  is not None else None
+
+        r,g,b = self.owner._map_channels_or_special(self.palette_name, ha, oo, si)
+
+        # match base size
+        H, W = self.base_rgb.shape[:2]
+        def fit(ch):
+            if ch is None: return None
+            if ch.shape[:2] != (H, W):
+                return self.owner._resize_to(ch, (W, H))
+            return ch
+        r, g, b = fit(r), fit(g), fit(b)
+
+        final = np.zeros_like(self.base_rgb, dtype=np.float32)
+        if r is not None: final[...,0] = r
+        if g is not None: final[...,1] = g
+        if b is not None: final[...,2] = b
+
+        m = float(final.max()) or 1.0
+        final = np.clip(final/m, 0.0, 1.0)
+
+        self.adjusted_image.emit(final)
+        self.accept()
+
+    def eventFilter(self, obj, evt):
+        if obj is self.preview_label:
+            if evt.type() == QEvent.Type.MouseButtonPress and evt.button() == Qt.MouseButton.LeftButton:
+                self._dragging = True; self._last_pos = evt.pos()
+                self.preview_label.setCursor(Qt.CursorShape.ClosedHandCursor); return True
+            if evt.type() == QEvent.Type.MouseMove and self._dragging:
+                d = evt.pos() - self._last_pos; self._last_pos = evt.pos()
+                self.preview_area.horizontalScrollBar().setValue(self.preview_area.horizontalScrollBar().value() - d.x())
+                self.preview_area.verticalScrollBar().setValue(self.preview_area.verticalScrollBar().value() - d.y())
+                return True
+            if evt.type() == QEvent.Type.MouseButtonRelease and evt.button() == Qt.MouseButton.LeftButton:
+                self._dragging = False; self.preview_label.setCursor(Qt.CursorShape.OpenHandCursor); return True
+            if evt.type() == QEvent.Type.Wheel:
+                self._change_zoom(1.1 if evt.angleDelta().y() > 0 else 0.9); return True
+        return super().eventFilter(obj, evt)
+
+
+class PerfectPalettePicker(QWidget):
+    THUMB_CROP = 512  # side length for thumbnail center crops
+    PALETTES = [
+        "SHO","HOO","HSO","HOS",
+        "OSS","OHH","OSH","OHS",
+        "HSS","Realistic1","Realistic2","Foraxx"
+    ]
+
+    def __init__(self, doc_manager=None, parent=None):
+        super().__init__(parent)
+        self.doc_manager = doc_manager
+        self.setWindowTitle("Perfect Palette Picker")
+
+        # raw channels (float32 ~[0..1])
+        self.ha   = None
+        self.oiii = None
+        self.sii  = None
+        self.osc1 = None
+        self.osc2 = None
+        self._dim_mismatch_accepted = False
+
+        # stretched cache (per input name → stretched array)
+        self._stretched: dict[str, np.ndarray] = {}
+
+        self.final = None
+        self.current_palette = None
+        self._thumb_base_pm: dict[str, QPixmap] = {}   # palette name -> base pixmap (image only)
+        self._selected_name: str | None = None
+
+        # thumbs
+        self._thumb_buttons: dict[str, QPushButton] = {}
+
+        self._base_pm: QPixmap | None = None
+        self._zoom = 1.0
+        self._min_zoom = 0.05
+        self._max_zoom = 6.0
+        self._panning = False
+        self._pan_last: QPoint | None = None
+
+        self._build_ui()
+
+    # ---------------- UI ----------------
+    def _build_ui(self):
+        root = QHBoxLayout(self)
+
+        # -------- left controls
+        left = QVBoxLayout()
+        left_host = QWidget(self); left_host.setLayout(left); left_host.setFixedWidth(300)
+
+        left.addWidget(QLabel("<b>Load channels</b>"))
+
+        # Load buttons + status labels
+        self.btn_ha   = QPushButton("Load Ha…");   self.btn_ha.clicked.connect(lambda: self._load_channel("Ha"))
+        self.btn_oiii = QPushButton("Load OIII…"); self.btn_oiii.clicked.connect(lambda: self._load_channel("OIII"))
+        self.btn_sii  = QPushButton("Load SII…");  self.btn_sii.clicked.connect(lambda: self._load_channel("SII"))
+        self.btn_osc1 = QPushButton("Load OSC1 (Ha/OIII)…"); self.btn_osc1.clicked.connect(lambda: self._load_channel("OSC1"))
+        self.btn_osc2 = QPushButton("Load OSC2 (SII/OIII)…"); self.btn_osc2.clicked.connect(lambda: self._load_channel("OSC2"))
+
+        self.lbl_ha   = QLabel("No Ha loaded.")
+        self.lbl_oiii = QLabel("No OIII loaded.")
+        self.lbl_sii  = QLabel("No SII loaded.")
+        self.lbl_osc1 = QLabel("No OSC1 loaded.")
+        self.lbl_osc2 = QLabel("No OSC2 loaded.")
+        for lab in (self.lbl_ha, self.lbl_oiii, self.lbl_sii, self.lbl_osc1, self.lbl_osc2):
+            lab.setWordWrap(True); lab.setStyleSheet("color:#888; margin-left:8px;")
+
+        for btn, lab in (
+            (self.btn_ha, self.lbl_ha),
+            (self.btn_oiii, self.lbl_oiii),
+            (self.btn_sii, self.lbl_sii),
+            (self.btn_osc1, self.lbl_osc1),
+            (self.btn_osc2, self.lbl_osc2),
+        ):
+            left.addWidget(btn); left.addWidget(lab)
+
+        # Linear toggle (stretch BEFORE palette build)
+        self.chk_linear = QCheckBox("Linear input (apply statistical stretch before build)")
+        self.chk_linear.setChecked(True)
+        self.chk_linear.stateChanged.connect(self._rebuild_stretch_cache_for_all)
+        left.addSpacing(6); left.addWidget(self.chk_linear)
+
+        # Actions
+        self.btn_clear = QPushButton("Clear Loaded Channels")
+        self.btn_clear.clicked.connect(self._clear_channels)
+        left.addWidget(self.btn_clear)
+
+        self.btn_create = QPushButton("Create Palettes")
+        self.btn_create.clicked.connect(self._create_palettes)
+        left.addWidget(self.btn_create)
+
+        self.btn_push = QPushButton("Push Final to New View")
+        self.btn_push.clicked.connect(self._push_final)
+        left.addWidget(self.btn_push)
+
+        left.addStretch(1)
+        root.addWidget(left_host, 0)
+
+        # -------- right: preview + fixed-size 4×3 grid
+        right = QVBoxLayout()
+
+        # zoom toolbar
+        # zoom toolbar (themed)
+        tools = QHBoxLayout()
+        self.btn_zoom_in  = themed_toolbtn("zoom-in", "Zoom In")
+        self.btn_zoom_out = themed_toolbtn("zoom-out", "Zoom Out")
+        self.btn_fit      = themed_toolbtn("zoom-fit-best", "Fit to Preview")
+
+        self.btn_zoom_in.clicked.connect(lambda: self._zoom_at(1.25))
+        self.btn_zoom_out.clicked.connect(lambda: self._zoom_at(0.8))
+        self.btn_fit.clicked.connect(self._fit_to_preview)
+
+        tools.addStretch(1)
+        tools.addWidget(self.btn_zoom_out)
+        tools.addWidget(self.btn_zoom_in)
+        tools.addWidget(self.btn_fit)
+        tools.addStretch(1)
+        right.addLayout(tools)
+
+
+        # main preview (expands)
+        self.scroll = QScrollArea(self); self.scroll.setWidgetResizable(True)
+        self.scroll.setAlignment(Qt.AlignmentFlag.AlignCenter)
+        self.preview = QLabel(alignment=Qt.AlignmentFlag.AlignCenter)
+        self.scroll.setWidget(self.preview)
+        self.preview.setMouseTracking(True)
+        self.preview.installEventFilter(self)
+        self.scroll.viewport().installEventFilter(self)     
+        self.scroll.installEventFilter(self)  
+        self.scroll.horizontalScrollBar().installEventFilter(self)  # NEW
+        self.scroll.verticalScrollBar().installEventFilter(self)    # NEW        
+        right.addWidget(self.scroll, 1)
+
+        # fixed-size grid
+        self.grid = QGridLayout()
+        self.grid.setHorizontalSpacing(8); self.grid.setVerticalSpacing(8)
+        self.grid.setContentsMargins(8, 8, 8, 8)
+
+        self.thumb_size = QSize(220, 110)
+        btn_w = self.thumb_size.width() + 2
+        btn_h = self.thumb_size.height() + 2
+        cols, rows = 4, 3
+
+        for idx, name in enumerate(self.PALETTES):
+            r, c = divmod(idx, cols)
+            b = QPushButton("")  # we draw the text onto the icon itself
+            b.setToolTip(name)
+            b.setIconSize(self.thumb_size)
+            b.setFixedSize(btn_w, btn_h)
+            b.setSizePolicy(QSizePolicy.Policy.Fixed, QSizePolicy.Policy.Fixed)
+            b.clicked.connect(lambda _=None, n=name: self._on_palette_clicked(n))
+            b.setStyleSheet("QPushButton{background:#222;border:1px solid #333;} QPushButton:hover{border-color:#555;}")
+            self._thumb_buttons[name] = b
+            self.grid.addWidget(b, r, c)
+
+        grid_host = QWidget(self); grid_host.setLayout(self.grid)
+        hspacing = self.grid.horizontalSpacing(); vspacing = self.grid.verticalSpacing()
+        m = self.grid.contentsMargins()
+        grid_w = cols*btn_w + (cols-1)*hspacing + m.left() + m.right()
+        grid_h = rows*btn_h + (rows-1)*vspacing + m.top() + m.bottom()
+        grid_host.setFixedSize(grid_w, grid_h)
+        grid_host.setSizePolicy(QSizePolicy.Policy.Fixed, QSizePolicy.Policy.Fixed)
+        right.addWidget(grid_host, 0, alignment=Qt.AlignmentFlag.AlignHCenter)
+
+        self.status = QLabel(""); right.addWidget(self.status, 0)
+
+        right_host = QWidget(self); right_host.setLayout(right)
+        root.addWidget(right_host, 1)
+
+        self.setLayout(root)
+        self.setMinimumSize(left_host.width() + grid_w + 48, max(560, grid_h + 200))
+
+    def _resize_to(self, arr: np.ndarray | None, size: tuple[int, int]) -> np.ndarray | None:
+        """Resize np array to (w,h). Keeps dtype/scale. Uses INTER_AREA for downsizing."""
+        if arr is None:
+            return None
+        w, h = size
+        if arr.ndim == 2:
+            src_h, src_w = arr.shape
+        else:
+            src_h, src_w = arr.shape[:2]
+        if (src_w, src_h) == (w, h):
+            return arr
+        if cv2 is None:
+            # ultra-simple fallback: nearest; OK for thumbs if OpenCV isn't present
+            if arr.ndim == 2:
+                return np.array(Image.fromarray((arr*255).astype(np.uint8)).resize((w, h))).astype(np.float32) / 255.0
+            return np.array(Image.fromarray((arr*255).astype(np.uint8)).resize((w, h))).astype(np.float32) / 255.0
+        interp = cv2.INTER_AREA if (w < src_w or h < src_h) else cv2.INTER_LINEAR
+        if arr.ndim == 2:
+            return cv2.resize(arr, (w, h), interpolation=interp)
+        return cv2.resize(arr, (w, h), interpolation=interp)
+
+    def _capture_view_state(self):
+        """Capture current view center in base-image coordinates + zoom."""
+        if self._base_pm is None:
+            return None
+        vp = self.scroll.viewport()
+        hbar = self.scroll.horizontalScrollBar()
+        vbar = self.scroll.verticalScrollBar()
+
+        # center of viewport in viewport coords
+        anchor_vp = QPoint(vp.width() // 2, vp.height() // 2)
+
+        # convert to label coords (scaled image coords)
+        anchor_lbl = self.preview.mapFrom(vp, anchor_vp)
+
+        # scaled -> base image coords
+        base_x = anchor_lbl.x() / max(self._zoom, 1e-6)
+        base_y = anchor_lbl.y() / max(self._zoom, 1e-6)
+
+        pm = self._base_pm.size()
+        fx = 0.5 if pm.width()  <= 0 else (base_x / pm.width())
+        fy = 0.5 if pm.height() <= 0 else (base_y / pm.height())
+
+        return {"zoom": float(self._zoom), "fx": float(fx), "fy": float(fy)}
+
+    def _restore_view_state(self, state):
+        """Restore zoom and pan using stored base-image fractions."""
+        if not state or self._base_pm is None:
+            return
+
+        # restore zoom first
+        self._zoom = max(self._min_zoom, min(self._max_zoom, float(state["zoom"])))
+        self._update_preview_pixmap()
+
+        # now restore center point
+        pm = self._base_pm.size()
+        fx = float(state.get("fx", 0.5))
+        fy = float(state.get("fy", 0.5))
+
+        base_x = fx * pm.width()
+        base_y = fy * pm.height()
+
+        # base -> scaled label coords
+        lbl_x = int(base_x * self._zoom)
+        lbl_y = int(base_y * self._zoom)
+
+        vp = self.scroll.viewport()
+        anchor_vp = QPoint(vp.width() // 2, vp.height() // 2)
+
+        hbar = self.scroll.horizontalScrollBar()
+        vbar = self.scroll.verticalScrollBar()
+        hbar.setValue(max(hbar.minimum(), min(hbar.maximum(), lbl_x - anchor_vp.x())))
+        vbar.setValue(max(vbar.minimum(), min(vbar.maximum(), lbl_y - anchor_vp.y())))
+
+    # ---------- status helpers ----------
+    def _set_status_label(self, which: str, text: str | None):
+        lab = getattr(self, f"lbl_{which.lower()}")
+        if text:
+            lab.setText(text)
+            lab.setStyleSheet("color:#2a7; font-weight:600; margin-left:8px;")
+        else:
+            lab.setText(f"No {which} loaded.")
+            lab.setStyleSheet("color:#888; margin-left:8px;")
+
+    # ------------- load by view/file -------------
+    def _load_channel(self, which: str):
+        src, ok = QInputDialog.getItem(
+            self, f"Load {which}", "Source:", ["From View", "From File"], 0, False
+        )
+        if not ok:
+            return
+
+        if src == "From View":
+            out = self._load_from_view(which)
+        else:
+            out = self._load_from_file(which)
+        if out is None:
+            return
+
+        img, header, bit_depth, is_mono, path, label = out
+
+        # NB channels → mono; OSC → RGB
+        if which in ("Ha","OIII","SII"):
+            if img.ndim == 3:
+                img = img[:, :, 0]
+        else:
+            if img.ndim == 2:
+                img = np.stack([img]*3, axis=-1)
+
+        # store raw, normalized
+        setattr(self, which.lower(), self._as_float01(img))
+        self._set_status_label(which, label)
+        self.status.setText(f"{which} loaded ({'mono' if img.ndim==2 else 'RGB'}) shape={img.shape}")
+
+        # build/clear stretched cache for this input
+        self._cache_stretch(which)
+
+        if self.current_palette is None:
+            self.current_palette = "SHO"
+
+    def _load_from_view(self, which):
+        views = self._list_open_views()
+        if not views:
+            QMessageBox.warning(self, "No Views", "No open image views were found.")
+            return None
+
+        labels = [lab for lab, _ in views]
+        choice, ok = QInputDialog.getItem(
+            self, f"Select View for {which}", "Choose a view (by name):", labels, 0, False
+        )
+        if not ok or not choice:
+            return None
+
+        sw = dict(views)[choice]
+        doc = getattr(sw, "document", None)
+        if doc is None or getattr(doc, "image", None) is None:
+            QMessageBox.warning(self, "Empty View", "Selected view has no image.")
+            return None
+
+        img = doc.image
+        meta = getattr(doc, "metadata", {}) or {}
+        header = meta.get("original_header", None)
+        bit_depth = meta.get("bit_depth", "Unknown")
+        is_mono = (img.ndim == 2) or (img.ndim == 3 and img.shape[2] == 1)
+        path = meta.get("file_path", None)
+        return img, header, bit_depth, is_mono, path, f"From View: {choice}"
+
+    def _load_from_file(self, which):
+        filt = "Images (*.png *.tif *.tiff *.fits *.fit *.xisf)"
+        path, _ = QFileDialog.getOpenFileName(self, f"Select {which} File", "", filt)
+        if not path:
+            return None
+        img, header, bit_depth, is_mono = legacy_load_image(path)
+        if img is None:
+            QMessageBox.critical(self, "Load Error", f"Could not load {os.path.basename(path)}")
+            return None
+        label = f"From File: {os.path.basename(path)}"
+        return img, header, bit_depth, is_mono, path, label
+
+    def showEvent(self, e):
+        super().showEvent(e)
+        QTimer.singleShot(0, self._center_scrollbars)
+
+    # ------------- build/caches -------------
+    def _cache_stretch(self, which: str):
+        """Compute and cache stretched version of a just-loaded input (if linear checked)."""
+        arr = getattr(self, which.lower())
+        if arr is None:
+            self._stretched.pop(which, None); return
+        if not self.chk_linear.isChecked():
+            self._stretched.pop(which, None); return
+        self._stretched[which] = self._stretch_input(arr)
+
+    def _rebuild_stretch_cache_for_all(self, _state: int):
+        """Rebuild (or clear) stretched cache for all loaded inputs when checkbox toggles."""
+        for which in ("Ha","OIII","SII","OSC1","OSC2"):
+            self._cache_stretch(which)
+
+    def _render_thumb(self, name: str):
+        base = self._thumb_base_pm.get(name)
+        if base is None:
+            return
+        pm = base.copy()
+
+        p = QPainter(pm)
+        p.setRenderHint(QPainter.RenderHint.Antialiasing)
+
+        font = QFont("Helvetica", 10, QFont.Weight.DemiBold)
+        p.setFont(font)
+        fm = QFontMetrics(font)
+
+        pad = 6
+        strip_h = fm.height() + pad * 2
+        strip = pm.rect().adjusted(0, pm.height() - strip_h, 0, 0)
+
+        # translucent bottom strip
+        p.fillRect(strip, QColor(0, 0, 0, 160))
+        color = QColor(102, 255, 102) if self._selected_name == name else QColor(255, 255, 255)
+        p.setPen(QPen(color))
+        p.drawText(strip, Qt.AlignmentFlag.AlignCenter, name)
+        p.end()
+
+        btn = self._thumb_buttons[name]
+        btn.setIcon(QIcon(pm))
+        btn.setIconSize(self.thumb_size)  # <- ensures no clipping
+
+    # ------------- thumbnails -------------
+    def _create_palettes(self):
+        """
+        Build the 12 palette thumbnails from a **center crop of the stretched inputs**
+        and draw the palette name directly on each thumbnail. Names turn green when selected.
+        """
+        ha, oo, si = self._prepared_channels(for_thumbs=True)
+        if oo is None or (ha is None and si is None):
+            QMessageBox.warning(self, "Need Channels", "Load at least OIII + (Ha or SII).")
+            return
+
+        built = 0
+        for name in self.PALETTES:
+            r, g, b = self._map_channels_or_special(name, ha, oo, si)
+            if any(ch is None for ch in (r, g, b)):
+                self._thumb_base_pm.pop(name, None)
+                self._thumb_buttons[name].setIcon(QIcon())
+                continue
+
+            r = np.clip(np.nan_to_num(r), 0, 1)
+            g = np.clip(np.nan_to_num(g), 0, 1)
+            b = np.clip(np.nan_to_num(b), 0, 1)
+            rgb = np.stack([r, g, b], axis=2).astype(np.float32)
+
+            # scale the thumbnail to EXACTLY the button icon size first
+            pm = QPixmap.fromImage(self._to_qimage(rgb)).scaled(
+                self.thumb_size, Qt.AspectRatioMode.KeepAspectRatio,
+                Qt.TransformationMode.SmoothTransformation
+            )
+            self._thumb_base_pm[name] = pm
+            self._render_thumb(name) 
+            built += 1
+
+        self.status.setText(f"Created {built} palette previews.")
+
+
+    def _on_palette_clicked(self, name: str):
+        self._selected_name = name
+        for n in self.PALETTES:
+            self._render_thumb(n)
+        self.current_palette = name
+        self._generate_for_palette(name)
+
+    # ------------- palette build helpers -------------
+    def _center_crop(self, img: np.ndarray, side: int) -> np.ndarray:
+        """Center-crop to a square of size 'side' (no upscaling)."""
+        h, w = img.shape[:2]; s = min(side, h, w)
+        y0 = (h - s) // 2; x0 = (w - s) // 2
+        return img[y0:y0+s, x0:x0+s] if img.ndim == 2 else img[y0:y0+s, x0:x0+s, :]
+
+    def _center_crop_all_to_side(self, side: int, *imgs):
+        """Center-crop all provided images to the same square side (no upscaling)."""
+        s = None
+        for im in imgs:
+            if im is None: continue
+            h, w = im.shape[:2]
+            s = min(side, h, w) if s is None else min(s, h, w, side)
+        if s is None: s = side
+        return [self._center_crop(im, s) if im is not None else None for im in imgs], s
+
+    def _prepared_channels(self, for_thumbs: bool = False):
+        """
+        Build Ha/OIII/SII bases from inputs. If 'Linear input' is checked,
+        **use stretched versions** (cached). Then optionally center-crop for thumbnails.
+        """
+        # choose raw vs stretched
+        def pick(name):
+            if self.chk_linear.isChecked() and (name in self._stretched):
+                return self._stretched[name]
+            return getattr(self, name.lower())
+
+        ha = pick("Ha")
+        oo = pick("OIII")
+        si = pick("SII")
+        o1 = pick("OSC1")
+        o2 = pick("OSC2")
+
+        # synthesize from stretched OSC first (stretch-before-crop)
+        if o1 is not None:  # OSC1: R≈Ha, mean(G,B)≈OIII
+            h1 = o1[..., 0]
+            g1b1 = o1[..., 1:3].mean(axis=2)
+            ha = h1 if ha is None else 0.5*ha + 0.5*h1
+            oo = g1b1 if oo is None else 0.5*oo + 0.5*g1b1
+
+        if o2 is not None:  # OSC2: R≈SII, mean(G,B)≈OIII
+            s2 = o2[..., 0]
+            g2b2 = o2[..., 1:3].mean(axis=2)
+            si = s2 if si is None else 0.5*si + 0.5*s2
+            oo = g2b2 if oo is None else 0.5*oo + 0.5*g2b2
+
+        # shapes must match for full-size
+        # shapes must match for full-size
+        shapes = [x.shape[:2] for x in (ha, oo, si) if x is not None]
+        if len(shapes) and len(set(shapes)) > 1 and not for_thumbs:
+            # pick a reference size (prefer Ha, then OIII, then SII)
+            ref = ha if ha is not None else (oo if oo is not None else si)
+            ref_name = "Ha" if ha is not None else ("OIII" if oo is not None else "SII")
+            ref_h, ref_w = ref.shape[:2]
+
+            # Only prompt once per session unless you want every time
+            if not self._dim_mismatch_accepted:
+                msg = (
+                    "The loaded channels have different image dimensions.\n\n"
+                    f"• Ha:   {None if ha is None else ha.shape}\n"
+                    f"• OIII: {None if oo is None else oo.shape}\n"
+                    f"• SII:  {None if si is None else si.shape}\n\n"
+                    f"SASpro can resize (warp) the channels to match the reference frame:\n"
+                    f"• Reference: {ref_name}\n"
+                    f"• Target size: ({ref_w} × {ref_h})\n\n"
+                    "Proceed and resize mismatched channels?"
+                )
+                ret = QMessageBox.question(
+                    self,
+                    "Channel Size Mismatch",
+                    msg,
+                    QMessageBox.StandardButton.Yes | QMessageBox.StandardButton.No,
+                    QMessageBox.StandardButton.Yes
+                )
+                if ret != QMessageBox.StandardButton.Yes:
+                    return None, None, None
+
+                self._dim_mismatch_accepted = True
+
+            # resize to reference
+            ha = self._resize_to(ha, (ref_w, ref_h)) if ha is not None else None
+            oo = self._resize_to(oo, (ref_w, ref_h)) if oo is not None else None
+            si = self._resize_to(si, (ref_w, ref_h)) if si is not None else None
+
+        # thumbnails: crop AFTER stretch/synth
+        if for_thumbs:
+            # choose a reference (prefer OIII, then Ha, then SII)
+            ref = oo if oo is not None else (ha if ha is not None else si)
+            if ref is not None:
+                ref_h, ref_w = ref.shape[:2]
+
+                # 1) first, size-match all channels to the reference full frame
+                ha = self._resize_to(ha, (ref_w, ref_h)) if ha is not None else None
+                oo = self._resize_to(oo, (ref_w, ref_h)) if oo is not None else None
+                si = self._resize_to(si, (ref_w, ref_h)) if si is not None else None
+
+                # 2) then, make a 50% view of the full rectangle
+                half_w = max(1, int(ref_w * 0.5))
+                half_h = max(1, int(ref_h * 0.5))
+                ha = self._resize_to(ha, (half_w, half_h)) if ha is not None else None
+                oo = self._resize_to(oo, (half_w, half_h)) if oo is not None else None
+                si = self._resize_to(si, (half_w, half_h)) if si is not None else None
+
+        return ha, oo, si
+
+    def _generate_for_palette(self, pal: str):
+        ha, oo, si = self._prepared_channels()
+        if oo is None or (ha is None and si is None):
+            return
+
+        r,g,b = self._map_channels_or_special(pal, ha, oo, si)
+        if any(ch is None for ch in (r,g,b)):
+            QMessageBox.critical(self, "Palette Error", f"Could not build palette {pal}."); return
+
+        r = np.clip(np.nan_to_num(r), 0, 1)
+        g = np.clip(np.nan_to_num(g), 0, 1)
+        b = np.clip(np.nan_to_num(b), 0, 1)
+        rgb = np.stack([r,g,b], axis=2).astype(np.float32)
+
+        mx = float(rgb.max()) or 1.0
+        self.final = (rgb / mx).astype(np.float32)
+
+        # Fit only when there wasn't an existing preview yet
+        first = (self._base_pm is None)
+        self._set_preview_image(self._to_qimage(self.final), fit=first, preserve_view=True)
+        self.status.setText(f"Preview generated: {pal}")
+
+    def _set_preview_image(self, qimg: QImage, *, fit: bool = False, preserve_view: bool = True):
+        state = None
+        if preserve_view and (not fit) and (self._base_pm is not None):
+            state = self._capture_view_state()
+
+        self._base_pm = QPixmap.fromImage(qimg)
+
+        # If we’re fitting, ignore old zoom/pan.
+        if fit or state is None:
+            self._zoom = 1.0
+            self._update_preview_pixmap()
+            if fit:
+                QTimer.singleShot(0, self._fit_to_preview)
+            else:
+                QTimer.singleShot(0, self._center_scrollbars)
+            return
+
+        # restore prior zoom/pan
+        self._restore_view_state(state)
+
+
+    def _update_preview_pixmap(self):
+        if self._base_pm is None:
+            return
+        # explicit int size (QSize * float can crash on some PyQt6 builds)
+        base_sz = self._base_pm.size()
+        w = max(1, int(base_sz.width() * self._zoom))
+        h = max(1, int(base_sz.height() * self._zoom))
+        scaled = self._base_pm.scaled(
+            w, h,
+            Qt.AspectRatioMode.KeepAspectRatio,
+            Qt.TransformationMode.SmoothTransformation
+        )
+        self.preview.setPixmap(scaled)
+        self.preview.resize(scaled.size())
+
+    def _set_zoom(self, new_zoom: float):
+        self._zoom = max(self._min_zoom, min(self._max_zoom, new_zoom))
+        self._update_preview_pixmap()
+
+    def _zoom_at(self, factor: float = 1.25, anchor_vp: QPoint | None = None):
+        if self._base_pm is None:
+            return
+
+        vp = self.scroll.viewport()
+        if anchor_vp is None:
+            anchor_vp = QPoint(vp.width() // 2, vp.height() // 2)  # view center
+
+        # label coords under the anchor *before* zoom
+        lbl_before = self.preview.mapFrom(vp, anchor_vp)
+
+        old_zoom = self._zoom
+        new_zoom = max(self._min_zoom, min(self._max_zoom, old_zoom * factor))
+        ratio = new_zoom / max(old_zoom, 1e-6)
+        if abs(ratio - 1.0) < 1e-6:
+            return
+
+        # apply zoom (updates label size & scrollbar ranges)
+        self._zoom = new_zoom
+        self._update_preview_pixmap()
+
+        # desired label coords *after* zoom
+        lbl_after_x = int(lbl_before.x() * ratio)
+        lbl_after_y = int(lbl_before.y() * ratio)
+
+        # move scrollbars so anchor_vp keeps the same content point
+        hbar = self.scroll.horizontalScrollBar()
+        vbar = self.scroll.verticalScrollBar()
+        hbar.setValue(max(hbar.minimum(), min(hbar.maximum(), lbl_after_x - anchor_vp.x())))
+        vbar.setValue(max(vbar.minimum(), min(vbar.maximum(), lbl_after_y - anchor_vp.y())))
+
+
+    def _fit_to_preview(self):
+        if self._base_pm is None:
+            return
+        vp = self.scroll.viewport().size()
+        pm = self._base_pm.size()
+        if pm.width() == 0 or pm.height() == 0:
+            return
+        k = min(vp.width() / pm.width(), vp.height() / pm.height())
+        self._set_zoom(max(self._min_zoom, min(self._max_zoom, k)))
+        self._center_scrollbars()
+
+    def _center_scrollbars(self):
+        # center the view on the image
+        h = self.scroll.horizontalScrollBar()
+        v = self.scroll.verticalScrollBar()
+        h.setValue((h.maximum() + h.minimum()) // 2)
+        v.setValue((v.maximum() + v.minimum()) // 2)
+
+    def _map_channels_or_special(self, name, ha, oo, si):
+        # substitution
+        if ha is None and si is not None: ha = si
+        if si is None and ha is not None: si = ha
+
+        basic = {
+            "SHO": (si, ha, oo),
+            "HOO": (ha, oo, oo),
+            "HSO": (ha, si, oo),
+            "HOS": (ha, oo, si),
+            "OSS": (oo, si, si),
+            "OHH": (oo, ha, ha),
+            "OSH": (oo, si, ha),
+            "OHS": (oo, ha, si),
+            "HSS": (ha, si, si),
+        }
+        if name in basic:
+            return basic[name]
+
+        try:
+            if name == "Realistic1":
+                r = (ha + si)/2 if (ha is not None and si is not None) else (ha if ha is not None else 0)
+                g = 0.3*(ha if ha is not None else 0) + 0.7*(oo if oo is not None else 0)
+                b = 0.9*(oo if oo is not None else 0) + 0.1*(ha if ha is not None else 0)
+                return r,g,b
+            if name == "Realistic2":
+                r = 0.7*(ha if ha is not None else 0) + 0.3*(si if si is not None else 0)
+                g = 0.3*(si if si is not None else 0) + 0.7*(oo if oo is not None else 0)
+                b = (oo if oo is not None else 0)
+                return r,g,b
+            if name == "Foraxx":
+                if ha is not None and oo is not None and si is None:
+                    r = ha; b = oo
+                    t = ha * oo
+                    g = (t**(1 - t))*ha + (1 - (t**(1 - t)))*oo
+                    return r,g,b
+                if ha is not None and oo is not None and si is not None:
+                    t = np.clip(oo, 1e-6, 1.0)**(1 - np.clip(oo, 1e-6, 1.0))
+                    r = t*si + (1 - t)*ha
+                    t2 = ha * oo
+                    g = (t2**(1 - t2))*ha + (1 - (t2**(1 - t2)))*oo
+                    b = oo
+                    return r,g,b
+                return basic["SHO"]
+        except Exception:
+            return basic.get("SHO", (ha, oo, si))
+
+        return basic.get("SHO", (ha, oo, si))
+
+    # ------------- push to new subwindow -------------
+    # ------------- push to new subwindow -------------
+    def _get_doc_manager(self):
+        """
+        Try several ways to get a DocManager:
+        1) explicit doc_manager passed into PerfectPalettePicker
+        2) main window's .docman or .doc_manager attribute
+        """
+        if self.doc_manager is not None:
+            return self.doc_manager
+
+        mw = self._find_main_window()
+        if mw is None:
+            return None
+
+        return getattr(mw, "docman", None) or getattr(mw, "doc_manager", None)
+
+    def _push_final(self):
+        if self.final is None:
+            QMessageBox.warning(self, "No Image", "Generate a palette first.")
+            return
+
+        # Use the SAME prepared channels the palette was built with
+        ha_prep, oo_prep, si_prep = self._prepared_channels()
+        if oo_prep is None or (ha_prep is None and si_prep is None):
+            QMessageBox.warning(self, "Need Channels", "Load at least OIII + (Ha or SII).")
+            return
+
+        dlg = PaletteAdjustDialog(
+            base_rgb     = self.final,                           # fully formed palette
+            palette_name = self.current_palette or "SHO",
+            ha_src       = ha_prep,                              # prepared (stretched/OSC-synth)
+            oiii_src     = oo_prep,
+            sii_src      = si_prep,
+            owner        = self
+        )
+        adjusted = {"img": None}
+        dlg.adjusted_image.connect(lambda img: adjusted.__setitem__("img", img))
+        dlg.exec()
+
+        if adjusted["img"] is None:
+            return  # user canceled
+
+        # Update preview with adjusted result and set as final
+        self.final = adjusted["img"]
+        self._set_preview_image(self._to_qimage(self.final))
+
+        title = self.current_palette or "Palette"
+
+        # ---- get DocManager the robust way ----
+        dm = self._get_doc_manager()
+
+        if dm is None:
+            # Fallback: open a simple viewer instead of erroring out
+            viewer = QDialog(self)
+            viewer.setWindowTitle(title)
+            vlayout = QVBoxLayout(viewer)
+            lbl = QLabel()
+            lbl.setAlignment(Qt.AlignmentFlag.AlignCenter)
+            lbl.setPixmap(QPixmap.fromImage(self._to_qimage(self.final)))
+            vlayout.addWidget(lbl)
+            viewer.resize(lbl.pixmap().size())
+            viewer.show()
+            # keep ref so it isn't GC'd
+            self._last_popup_viewer = viewer
+            self.status.setText("DocManager not found; opened palette in stand-alone viewer.")
+            return
+
+        # ---- normal SAS path: create a new document ----
+        try:
+            if hasattr(dm, "open_array"):
+                # many of your tools already use this signature
+                doc = dm.open_array(self.final, metadata={"is_mono": False}, title=title)
+            elif hasattr(dm, "create_document"):
+                doc = dm.create_document(image=self.final, metadata={"is_mono": False}, name=title)
+            else:
+                raise RuntimeError("DocManager lacks open_array/create_document")
+
+            # If DocManager or main window auto-spawns subwindows on new docs,
+            # this is all we need. If not, you can optionally keep the
+            # _spawn_subwindow_for hook here.
+            self.status.setText("Opened final palette in a new view.")
+        except Exception as e:
+            QMessageBox.critical(self, "Error", f"Failed to open new view:\n{e}")
+
+
+
+    # ------------- utilities -------------
+    def _clear_channels(self):
+        self.ha = self.oiii = self.sii = self.osc1 = self.osc2 = None
+        self._stretched.clear()
+        self._dim_mismatch_accepted = False
+        self.final = None
+        self.preview.clear()
+        for which in ("Ha","OIII","SII","OSC1","OSC2"):
+            self._set_status_label(which, None)
+        for name, b in self._thumb_buttons.items():
+            b.setIcon(QIcon())
+        self._thumb_base_pm.clear()
+        self._selected_name = None
+        for b in self._thumb_buttons.values():
+            b.setIcon(QIcon())
+        self.status.setText("Cleared all loaded channels.")
+
+    def _as_float01(self, arr):
+        a = np.asarray(arr)
+        if a.dtype == np.uint8:  return a.astype(np.float32)/255.0
+        if a.dtype == np.uint16: return a.astype(np.float32)/65535.0
+        return np.clip(a.astype(np.float32), 0.0, 1.0)
+
+    def _stretch_input(self, img):
+        """Run statistical stretch on mono or color inputs (target_median=0.25)."""
+        if img.ndim == 2:
+            return np.clip(stretch_mono_image(img, target_median=0.25), 0.0, 1.0)
+        if img.ndim == 3 and img.shape[2] == 3:
+            return np.clip(stretch_color_image(img, target_median=0.25, linked=False), 0.0, 1.0)
+        if img.ndim == 3 and img.shape[2] == 1:
+            mono = img[...,0]
+            return np.clip(stretch_mono_image(mono, target_median=0.25), 0.0, 1.0)
+        return img
+
+    def _to_qimage(self, arr):
+        a = np.clip(arr, 0, 1)
+        if a.ndim == 2:
+            u = (a * 255).astype(np.uint8); h, w = u.shape
+            return QImage(u.data, w, h, w, QImage.Format.Format_Grayscale8).copy()
+        if a.ndim == 3 and a.shape[2] == 3:
+            u = (a * 255).astype(np.uint8); h, w, _ = u.shape
+            return QImage(u.data, w, h, w*3, QImage.Format.Format_RGB888).copy()
+        raise ValueError(f"Unexpected image shape: {a.shape}")
+
+    def _find_main_window(self):
+        w = self
+        from PyQt6.QtWidgets import QMainWindow, QApplication
+        while w is not None and not isinstance(w, QMainWindow):
+            w = w.parentWidget()
+        if w: return w
+        for tlw in QApplication.topLevelWidgets():
+            if isinstance(tlw, QMainWindow):
+                return tlw
+        return None
+
+    def _list_open_views(self):
+        mw = self._find_main_window()
+        if not mw:
+            return []
+        try:
+            from setiastro.saspro.subwindow import ImageSubWindow
+            subs = mw.findChildren(ImageSubWindow)
+        except Exception:
+            subs = []
+        out = []
+        for sw in subs:
+            title = getattr(sw, "view_title", None) or sw.windowTitle() or getattr(sw.document, "display_name", lambda: "Untitled")()
+            out.append((str(title), sw))
+        return out
+    
+    def eventFilter(self, obj, ev):
+        # Ctrl+wheel = zoom at mouse (no scrolling). Wheel without Ctrl = eaten.
+        if ev.type() == QEvent.Type.Wheel and (
+            obj is self.preview
+            or obj is self.scroll
+            or obj is self.scroll.viewport()
+            or obj is self.scroll.horizontalScrollBar()
+            or obj is self.scroll.verticalScrollBar()
+        ):
+            # always stop the wheel from scrolling
+            ev.accept()
+
+            # Zoom only when Ctrl is held
+            if ev.modifiers() & Qt.KeyboardModifier.ControlModifier:
+                factor = 1.25 if ev.angleDelta().y() > 0 else 0.8
+
+                # Get mouse position in global screen coords and map into the viewport
+                vp = self.scroll.viewport()
+                anchor_vp = vp.mapFromGlobal(ev.globalPosition().toPoint())
+
+                # Clamp to viewport rect (robust if the event originated on scrollbars)
+                r = vp.rect()
+                if not r.contains(anchor_vp):
+                    anchor_vp.setX(max(r.left(),  min(r.right(),  anchor_vp.x())))
+                    anchor_vp.setY(max(r.top(),   min(r.bottom(), anchor_vp.y())))
+
+                self._zoom_at(factor, anchor_vp)
+            return True
+        # click-drag pan on viewport
+        if obj is self.scroll.viewport():
+            if ev.type() == QEvent.Type.MouseButtonPress and ev.button() == Qt.MouseButton.LeftButton:
+                self._panning = True
+                self._pan_last = ev.position().toPoint()
+                self.scroll.viewport().setCursor(QCursor(Qt.CursorShape.ClosedHandCursor))
+                return True
+            if ev.type() == QEvent.Type.MouseMove and self._panning:
+                cur = ev.position().toPoint()
+                delta = cur - (self._pan_last or cur)
+                self._pan_last = cur
+                h = self.scroll.horizontalScrollBar()
+                v = self.scroll.verticalScrollBar()
+                h.setValue(h.value() - delta.x())
+                v.setValue(v.value() - delta.y())
+                return True
+            if ev.type() == QEvent.Type.MouseButtonRelease and ev.button() == Qt.MouseButton.LeftButton:
+                self._panning = False
+                self._pan_last = None
+                self.scroll.viewport().setCursor(QCursor(Qt.CursorShape.ArrowCursor))
+                return True
+
+        return super().eventFilter(obj, ev)