setiastrosuitepro 1.6.5.post3__py3-none-any.whl

setiastro/saspro/histogram.py

@@ -0,0 +1,756 @@
+# pro/histogram.py
+from __future__ import annotations
+import numpy as np
+
+from PyQt6.QtCore import Qt, QSettings, QTimer, QEvent, pyqtSignal
+from PyQt6.QtWidgets import (
+    QDialog, QVBoxLayout, QHBoxLayout, QLabel, QSlider, QPushButton, QScrollArea,
+    QTableWidget, QTableWidgetItem, QMessageBox, QToolButton, QInputDialog, QSplitter, QSizePolicy, QHeaderView
+)
+from PyQt6.QtGui import QPixmap, QPainter, QPen, QColor, QFont, QPalette
+
+# Shared utilities
+from setiastro.saspro.widgets.image_utils import to_float01 as _to_float01
+from setiastro.saspro.widgets.themed_buttons import themed_toolbtn
+
+def _to_float_preserve(img):
+    if img is None: return None
+    a = np.asarray(img)
+    return a.astype(np.float32, copy=False) if a.dtype != np.float32 else a
+
+
+
+class HistogramDialog(QDialog):
+    """
+    Per-document histogram (non-modal).
+    - Connects to ImageDocument.changed and repaints automatically.
+    - Multiple dialogs can be open at once (each bound to one doc).
+    """
+    pivotPicked = pyqtSignal(float)  # normalized [0..1] x position for GHS pivot
+    def __init__(self, parent, document):
+        super().__init__(parent)
+        self.setWindowTitle(self.tr("Histogram"))
+        self.setWindowFlag(Qt.WindowType.Window, True)
+        self.setWindowModality(Qt.WindowModality.NonModal)
+        self.setModal(False)
+        self.doc = document
+        self.image = _to_float_preserve(document.image)
+
+        self.zoom_factor = 1.0   # 1.0 = 100%
+        self.log_scale   = False # log X
+        self.log_y       = False # log Y
+        self.setAttribute(Qt.WidgetAttribute.WA_DeleteOnClose, True)
+        self._eps_log         = 1e-6      # first log bin edge (for labels)
+
+        # for mapping clicks → normalized x
+        self._click_mapping = None  # dict or None
+        self.settings = QSettings()
+        self.sensor_max01 = 1.0
+        self.sensor_native_max = None     # user ADU max (e.g., 65532)
+        self.native_theoretical_max = None
+
+        # histogram cache
+        self._bin_count       = 512
+        self._bin_edges_lin   = None      # np.ndarray | None
+        self._bin_edges_log   = None      # np.ndarray | None
+        self._counts_lin      = None      # list[np.ndarray] | None
+        self._counts_log      = None      # list[np.ndarray] | None
+        self._is_color        = False
+        self._eps_log         = 1e-6      # first log bin edge (for labels)
+
+        self._load_sensor_max_setting()
+        self._build_ui()
+
+        # debounce timer for resize / splitter moves
+        self._resize_timer = QTimer(self)
+        self._resize_timer.setSingleShot(True)
+        self._resize_timer.setInterval(80)  # ms; tweak if you want snappier/slower
+        self._resize_timer.timeout.connect(self._draw_histogram)
+
+        # prime histogram & stats from initial image
+        self._recompute_hist_cache()
+        self._update_stats()
+
+        # wire up to this specific document
+        self.doc.changed.connect(self._on_doc_changed)
+        # If the doc object goes away, close this dialog
+        self.doc.destroyed.connect(self.deleteLater)
+
+        self.setAttribute(Qt.WidgetAttribute.WA_DeleteOnClose, True)
+        self._doc_conn = False
+        if getattr(self, "doc", None) is not None:
+            try:
+                self.doc.destroyed.connect(self._on_doc_destroyed)
+                self._doc_conn = True
+            except Exception:
+                pass
+
+        # Do the first draw once the widget has a real size
+        QTimer.singleShot(0, self._draw_histogram)
+
+
+
+    # ---------- UI ----------
+    def _build_ui(self):
+        # Make it start at a sensible size
+        self.setMinimumSize(800, 400)
+        self.resize(900, 500)
+
+        main_layout = QVBoxLayout(self)
+
+        # --- top area: splitter with histogram + stats ---
+        splitter = QSplitter(Qt.Orientation.Horizontal, self)
+
+        # left: scroll area + label for the pixmap
+        self.scroll_area = QScrollArea(self)
+        self.scroll_area.setWidgetResizable(True)
+        self.scroll_area.setSizePolicy(
+            QSizePolicy.Policy.Expanding,
+            QSizePolicy.Policy.Expanding,
+        )
+
+        self.hist_label = QLabel(self)
+        self.hist_label.setAlignment(Qt.AlignmentFlag.AlignCenter)
+        self.scroll_area.setWidget(self.hist_label)
+        self.hist_label.installEventFilter(self)
+        self.hist_label.setToolTip(self.tr(
+            "Ctrl+Click on the histogram to send that intensity as the "
+            "pivot to Hyperbolic Stretch (if open)."
+        ))
+        self.scroll_area.viewport().installEventFilter(self)
+
+        splitter.addWidget(self.scroll_area)
+
+        # right: stats table
+        self.stats_table = QTableWidget(self)
+        self.stats_table.setRowCount(7)
+        self.stats_table.setColumnCount(1)
+        self.stats_table.setVerticalHeaderLabels([
+            self.tr("Min"), self.tr("Max"), self.tr("Median"), self.tr("StdDev"),
+            self.tr("MAD"), self.tr("Low Clipped"), self.tr("High Clipped")
+        ])
+
+        # Let it grow/shrink with the splitter
+        self.stats_table.setMinimumWidth(320)
+        self.stats_table.setSizePolicy(
+            QSizePolicy.Policy.Preferred,      # <- was Fixed
+            QSizePolicy.Policy.Expanding,
+        )
+
+        # Make the columns use available width nicely
+        hdr = self.stats_table.horizontalHeader()
+        hdr.setSectionResizeMode(QHeaderView.ResizeMode.Stretch)
+        # hdr.setStretchLastSection(True)
+        splitter.addWidget(self.stats_table)
+
+        # Give more space to histogram side by default
+        splitter.setStretchFactor(0, 3)
+        splitter.setStretchFactor(1, 1)
+        # Explicit initial sizes so it doesn't start with a tiny histogram
+        splitter.setSizes([650, 250])
+
+        QTimer.singleShot(0, self._adjust_stats_width)
+
+        main_layout.addWidget(splitter)
+
+        # --- controls row (unchanged except for being below splitter) ---
+        ctl = QHBoxLayout()
+        self.zoom_slider = QSlider(Qt.Orientation.Horizontal, self)
+        self.zoom_slider.setRange(50, 1000)
+        self.zoom_slider.setValue(100)
+        self.zoom_slider.setTickInterval(10)
+        self.zoom_slider.setTickPosition(QSlider.TickPosition.TicksBelow)
+        self.zoom_slider.valueChanged.connect(self._on_zoom_changed)
+
+        ctl.addWidget(QLabel(self.tr("Zoom:")))
+        ctl.addWidget(self.zoom_slider)
+
+        self.btn_logx = QPushButton(self.tr("Toggle Log X-Axis"), self)
+        self.btn_logx.setCheckable(True)
+        self.btn_logx.toggled.connect(self._toggle_log_x)
+        ctl.addWidget(self.btn_logx)
+
+        self.btn_logy = QPushButton(self.tr("Toggle Log Y-Axis"), self)
+        self.btn_logy.setCheckable(True)
+        self.btn_logy.toggled.connect(self._toggle_log_y)
+        ctl.addWidget(self.btn_logy)
+
+        self.btn_sensor_max = QToolButton(self)
+        self.btn_sensor_max.setText("?")
+        self.btn_sensor_max.setToolTip(self.tr(
+            "Set your camera's true saturation level for clipping warnings.\n"
+            "Tip: take an overexposed frame and see its max ADU."
+        ))
+        self.btn_sensor_max.clicked.connect(self._prompt_sensor_max)
+        ctl.addWidget(self.btn_sensor_max)
+
+        main_layout.addLayout(ctl)
+
+        btn_close = QPushButton(self.tr("Close"), self)
+        btn_close.clicked.connect(self.accept)
+        main_layout.addWidget(btn_close)
+
+        self.setLayout(main_layout)
+
+
+
+    # ---------- slots ----------
+    def _on_doc_changed(self):
+        self.image = _to_float_preserve(self.doc.image)
+        self._recompute_hist_cache()
+        self._update_stats()
+        self._draw_histogram()
+
+    def _on_zoom_changed(self, v: int):
+        self.zoom_factor = v / 100.0
+        self._draw_histogram()
+
+    def _toggle_log_x(self, on: bool):
+        self.log_scale = bool(on)
+        self._draw_histogram()
+
+    def _toggle_log_y(self, on: bool):
+        self.log_y = bool(on)
+        self._draw_histogram()
+
+    # ---------- drawing ----------
+    # ---------- drawing ----------
+    def _draw_histogram(self):
+        # nothing to draw yet
+        if self.image is None or self._bin_edges_lin is None:
+            self.hist_label.clear()
+            return
+
+        # use available size in the scroll area's viewport
+        if self.scroll_area is not None:
+            vp = self.scroll_area.viewport()
+            avail_w = max(200, vp.width())
+            avail_h = max(200, vp.height())
+        else:
+            avail_w = 512
+            avail_h = 300
+
+        base_width = avail_w
+        height = avail_h
+        width = int(base_width * self.zoom_factor)
+
+        # layout margins
+        left_margin    = 32   # room for Y labels
+        top_margin     = 12   # room so top ticks/text aren't clipped
+        bottom_margin  = 24   # room for X labels
+        axis_y         = height - bottom_margin
+        usable_h       = max(1, axis_y - top_margin)
+        plot_width     = max(1, width - left_margin)
+
+        # choose edges + raw counts from cache
+        if self.log_scale:
+            bin_edges   = self._bin_edges_log
+            counts_list = self._counts_log
+        else:
+            bin_edges   = self._bin_edges_lin
+            counts_list = self._counts_lin
+
+        if bin_edges is None or counts_list is None:
+            self.hist_label.clear()
+            return
+
+        bin_count = len(bin_edges) - 1
+
+        # precompute log range if needed
+        if self.log_scale:
+            # guard: avoid log10(<=0)
+            be0 = float(bin_edges[0])
+            if be0 <= 0:
+                be0 = self._eps_log
+            log_min = np.log10(be0)
+            log_max = 0.0
+        else:
+            log_min = None
+            log_max = None
+
+        # map X-domain edge → pixel X
+        def x_pos(edge: float) -> int:
+            if self.log_scale:
+                if edge <= 0:
+                    edge = self._eps_log
+                if abs(log_max - log_min) < 1e-12:
+                    return left_margin
+                return left_margin + int(
+                    (np.log10(edge) - log_min) / (log_max - log_min) * plot_width
+                )
+            else:
+                return left_margin + int(edge * plot_width)
+
+        # --- convert counts → display values (linear or log Y) ---
+        vals_list: list[np.ndarray] = []
+        max_val = 0.0
+        for counts in counts_list:
+            if self.log_y:
+                vals = np.log10(counts + 1.0)
+            else:
+                vals = counts.astype(np.float32)
+            if vals.size:
+                max_val = max(max_val, float(vals.max()))
+            vals_list.append(vals)
+
+        if max_val <= 0:
+            max_val = 1.0
+
+        # theme colors
+        pal = self.window().palette() if self.window() else self.palette()
+        bg_color   = pal.color(QPalette.ColorRole.Window)
+        text_color = pal.color(QPalette.ColorRole.Text)
+
+        if bg_color.lightness() < 128:
+            axis_color  = QColor(210, 210, 210)
+            label_color = QColor(245, 245, 245)
+        else:
+            axis_color  = QColor(40, 40, 40)
+            label_color = text_color
+
+        grid_color = QColor(axis_color)
+        grid_color.setAlpha(60)
+        grid_pen = QPen(grid_color)
+        grid_pen.setWidth(1)
+
+        pm = QPixmap(width, height)
+        pm.fill(bg_color)
+        p = QPainter(pm)
+        p.setRenderHint(QPainter.RenderHint.Antialiasing)
+
+        # helper: map normalized [0,1] → Y pixel (0 at bottom, 1 at top)
+        def y_pos(norm: float) -> int:
+            # norm in [0,1], map 0→axis_y, 1→top_margin
+            return int(top_margin + (1.0 - norm) * usable_h)
+
+        # ----- draw bars -----
+        if self._is_color:
+            colors = [
+                QColor(255, 0, 0, 140),
+                QColor(0, 180, 0, 140),
+                QColor(0, 0, 255, 140),
+            ]
+            for ch_idx, vals in enumerate(vals_list):
+                hn = vals / max_val
+                p.setPen(QPen(colors[ch_idx]))
+                for i in range(bin_count):
+                    x0 = x_pos(float(bin_edges[i]))
+                    x1 = x_pos(float(bin_edges[i + 1]))
+                    w  = max(1, x1 - x0)
+                    h  = int(hn[i] * usable_h)
+                    y0 = axis_y - h
+                    p.drawRect(x0, y0, w, h)
+        else:
+            vals = vals_list[0]
+            hn = vals / max_val
+            p.setPen(QPen(axis_color))
+            for i in range(bin_count):
+                x0 = x_pos(float(bin_edges[i]))
+                x1 = x_pos(float(bin_edges[i + 1]))
+                w  = max(1, x1 - x0)
+                h  = int(hn[i] * usable_h)
+                y0 = axis_y - h
+                p.drawRect(x0, y0, w, h)
+
+        # ----- axes -----
+        p.setPen(QPen(axis_color, 2))
+        # X axis at axis_y, Y axis from top_margin down to axis_y
+        p.drawLine(left_margin, axis_y, width - 1, axis_y)
+        p.drawLine(left_margin, top_margin, left_margin, axis_y)
+
+        p.setFont(QFont("Arial", 10))
+
+        # ----- X ticks + grid -----
+        if self.log_scale:
+            ticks = np.logspace(np.log10(bin_edges[0]), 0.0, 11)
+            for t in ticks:
+                x = x_pos(float(t))
+                if left_margin < x < width - 1:
+                    p.setPen(grid_pen)
+                    p.drawLine(x, top_margin, x, axis_y)
+                p.setPen(axis_color)
+                p.drawLine(x, axis_y, x, axis_y - 5)
+                p.setPen(label_color)
+                p.drawText(x - 18, axis_y + bottom_margin - 8, f"{t:.3f}")
+        else:
+            ticks = np.linspace(0.0, 1.0, 11)
+            for t in ticks:
+                x = x_pos(float(t))
+                if left_margin < x < width - 1:
+                    p.setPen(grid_pen)
+                    p.drawLine(x, top_margin, x, axis_y)
+                p.setPen(axis_color)
+                p.drawLine(x, axis_y, x, axis_y - 5)
+                p.setPen(label_color)
+                p.drawText(x - 10, axis_y + bottom_margin - 8, f"{t:.1f}")
+
+        # ----- Y ticks + grid -----
+        n_yticks = 6
+        if self.log_y:
+            exps = np.linspace(0.0, max_val, n_yticks)
+            norms = exps / max_val
+            labels = [f"{10**e:.0f}" for e in exps]
+        else:
+            vals_for_ticks = np.linspace(0.0, max_val, n_yticks)
+            norms = vals_for_ticks / max_val
+            labels = [f"{v:.0f}" for v in vals_for_ticks]
+
+        for i, (yn, lab) in enumerate(zip(norms, labels)):
+            y = y_pos(float(yn))
+            if 0 < i < n_yticks - 1:
+                p.setPen(grid_pen)
+                p.drawLine(left_margin, y, width - 1, y)
+            p.setPen(axis_color)
+            p.drawLine(left_margin - 5, y, left_margin, y)
+            p.setPen(label_color)
+            p.drawText(2, y + 4, lab)
+
+        # --- draw effective-max marker if user set one ---
+        if self.sensor_max01 < 0.9999:
+            x = x_pos(self.sensor_max01)
+            p.setPen(QPen(QColor(220, 0, 0), 2, Qt.PenStyle.DashLine))
+            p.drawLine(x, top_margin, x, axis_y)
+            p.drawText(min(x + 4, width - 80), top_margin + 12,
+                       self.tr("True Max {0:.4f}").format(self.sensor_max01))
+        # store mapping info for Ctrl+click → normalized x
+        try:
+            self._click_mapping = {
+                "left_margin": left_margin,
+                "plot_width": plot_width,
+                "axis_y": axis_y,
+                "top_margin": top_margin,
+                "height": height,
+                "log_scale": bool(self.log_scale),
+                "log_min": log_min,
+                "log_max": log_max,
+            }
+        except Exception:
+            self._click_mapping = None
+        p.end()
+        self.hist_label.setPixmap(pm)
+        self.hist_label.resize(pm.size())
+
+    def _x_pix_to_u(self, x_pix: int) -> float | None:
+        """
+        Map a horizontal pixel coordinate (in the label) to a normalized
+        intensity in [0..1], respecting linear / log X modes.
+        """
+        m = self._click_mapping
+        if not m:
+            return None
+
+        left = m["left_margin"]
+        width = max(1, m["plot_width"])
+        if x_pix < left or x_pix > left + width:
+            return None
+
+        t = (x_pix - left) / float(width)
+        t = max(0.0, min(1.0, t))
+
+        if not m["log_scale"]:
+            # linear: domain is already [0..1]
+            return float(t)
+
+        # log X: t in [0..1] corresponds to [10^log_min .. 10^log_max] (log_max ~ 0)
+        log_min = m.get("log_min", None)
+        log_max = m.get("log_max", None)
+        if log_min is None or log_max is None or abs(log_max - log_min) < 1e-12:
+            return float(t)
+
+        log_v = log_min + t * (log_max - log_min)
+        v = 10.0 ** log_v
+        # v is in (eps .. 1]; clamp to [0..1]
+        return float(max(0.0, min(1.0, v)))
+
+
+    def _recompute_hist_cache(self):
+        """Compute histograms once for the current image.
+
+        This is called when the document image changes. Resizing / zooming
+        will only redraw using this cached data.
+        """
+        img = self.image
+        self._bin_edges_lin = None
+        self._bin_edges_log = None
+        self._counts_lin    = None
+        self._counts_log    = None
+        self._is_color      = False
+        self._eps_log       = 1e-6
+
+        if img is None:
+            return
+
+        a = img
+        if a.ndim == 3 and a.shape[2] == 1:
+            a = a[..., 0]
+
+        if a.ndim == 3 and a.shape[2] == 3:
+            chans = [a[..., i] for i in range(3)]
+            self._is_color = True
+        else:
+            chan = a if a.ndim == 2 else a[..., 0]
+            chans = [chan]
+            self._is_color = False
+
+        bin_count = self._bin_count
+
+        # --- linear X bins ---
+        bin_edges_lin = np.linspace(0.0, 1.0, bin_count + 1).astype(np.float32)
+        counts_lin: list[np.ndarray] = []
+        for c in chans:
+            counts, _ = np.histogram(c.ravel(), bins=bin_edges_lin)
+            counts_lin.append(counts.astype(np.float32))
+
+        # --- log X bins ---
+        pos = a[a > 0]
+        eps = max(1e-6, float(pos.min())) if pos.size else 1e-6
+        log_min, log_max = np.log10(eps), 0.0
+        if abs(log_max - log_min) < 1e-12:
+            bin_edges_log = np.linspace(eps, 1.0, bin_count + 1).astype(np.float32)
+        else:
+            bin_edges_log = np.logspace(log_min, log_max, bin_count + 1).astype(np.float32)
+
+        counts_log: list[np.ndarray] = []
+        for c in chans:
+            counts, _ = np.histogram(c.ravel(), bins=bin_edges_log)
+            counts_log.append(counts.astype(np.float32))
+
+        self._bin_edges_lin = bin_edges_lin
+        self._bin_edges_log = bin_edges_log
+        self._counts_lin    = counts_lin
+        self._counts_log    = counts_log
+        self._eps_log       = float(eps)
+
+
+    def _schedule_redraw(self):
+        # Only bother if visible; restart timer each time
+        if self.isVisible():
+            self._resize_timer.start()
+
+    def resizeEvent(self, event):
+        super().resizeEvent(event)
+        self._schedule_redraw()
+
+    def eventFilter(self, obj, event):
+        # Ctrl+click on the histogram pixmap → emit pivotPicked(u)
+        if obj is self.hist_label and event.type() == QEvent.Type.MouseButtonPress:
+            if (event.button() == Qt.MouseButton.LeftButton and
+                    (event.modifiers() & Qt.KeyboardModifier.ControlModifier)):
+                pos = event.position().toPoint()
+                u = self._x_pix_to_u(pos.x())
+                if u is not None:
+                    # emit normalized pivot in [0..1]
+                    self.pivotPicked.emit(u)
+                    event.accept()
+                    return True
+
+        # When the splitter moves, the scroll_area viewport gets a Resize event
+        if self.scroll_area is not None and obj is self.scroll_area.viewport():
+            if event.type() == QEvent.Type.Resize:
+                self._schedule_redraw()
+
+        return super().eventFilter(obj, event)
+
+    def _update_stats(self):
+        if self.image is None:
+            return
+
+        img = self.image
+        # determine channels
+        if img.ndim == 3 and img.shape[2] == 3:
+            chans = [img[..., i] for i in range(3)]
+            self.stats_table.setColumnCount(3)
+            self.stats_table.setHorizontalHeaderLabels(["R", "G", "B"])
+        else:
+            chan = img if img.ndim == 2 else img[..., 0]
+            chans = [chan]
+            self.stats_table.setColumnCount(1)
+            self.stats_table.setHorizontalHeaderLabels(["Gray"])
+
+        eps = 1e-6  # tolerance for "exactly 0/1" after float ops
+
+        row_defs = [
+            (self.tr("Min"),          lambda c: float(np.min(c)),                "{:.4f}"),
+            (self.tr("Max"),          lambda c: float(np.max(c)),                "{:.4f}"),
+            (self.tr("Median"),       lambda c: float(np.median(c)),             "{:.4f}"),
+            (self.tr("StdDev"),       lambda c: float(np.std(c)),                "{:.4f}"),
+            (self.tr("MAD"),          lambda c: float(np.median(np.abs(c - np.median(c)))), "{:.4f}"),
+            (self.tr("Low Clipped"),  lambda c: _clip_fmt(c, low=True,  eps=eps), "{}"),
+            (self.tr("High Clipped"), lambda c: _clip_fmt(c, low=False, eps=eps), "{}"),
+        ]
+
+        def _clip_fmt(c, low: bool, eps: float):
+            flat = np.ravel(c)
+            n = flat.size if flat.size else 1
+            if low:
+                k = int(np.count_nonzero(flat <= eps))
+            else:
+                hi_thr = max(eps, self.sensor_max01 - eps)
+                k = int(np.count_nonzero(flat >= hi_thr))
+            pct = 100.0 * k / n
+            return f"{k} ({pct:.3f}%)"
+
+        # apply labels + sizes
+        self.stats_table.setRowCount(len(row_defs))
+        self.stats_table.setVerticalHeaderLabels([lab for lab, _, _ in row_defs])
+
+        # fill cells
+        for r, (lab, fn, fmt) in enumerate(row_defs):
+            for c_idx, c_arr in enumerate(chans):
+                val = fn(c_arr)
+                text = fmt.format(val)
+                it = QTableWidgetItem(text)
+                it.setTextAlignment(Qt.AlignmentFlag.AlignCenter)
+
+                # --- visual pop for non-trivial clipping ---
+                if lab in (self.tr("Low Clipped"), self.tr("High Clipped")):
+                    # text looks like: "123 (0.456%)"
+                    try:
+                        pct_str = text.split("(")[1].split("%")[0]
+                        pct = float(pct_str)
+                    except Exception:
+                        pct = 0.0
+
+                    # thresholds you can tweak
+                    # <0.01%: ignore
+                    # 0.01–0.1%: mild warning
+                    # 0.1–1%: clear warning
+                    # >1%: strong warning
+                    if pct >= 1.0:
+                        it.setBackground(QColor(100, 30, 30))  # strong red tint
+                    elif pct >= 0.1:
+                        it.setBackground(QColor(70, 30, 30))  # medium red tint
+                    elif pct >= 0.01:
+                        it.setBackground(QColor(40, 30, 30))  # mild red tint
+
+                self.stats_table.setItem(r, c_idx, it)
+
+        self._adjust_stats_width()        
+
+    def _theoretical_native_max_from_meta(self):
+        meta = getattr(self.doc, "metadata", None) or {}
+        bd = str(meta.get("bit_depth", "")).lower()
+
+        if "16-bit" in bd:
+            return 65535
+        if "8-bit" in bd:
+            return 255
+        if "32-bit unsigned" in bd:
+            return 4294967295
+        return None
+
+    def _settings_key_for_native_max(self, native_theoretical_max):
+        if native_theoretical_max == 65535:
+            return "histogram/sensor_max_native_16"
+        if native_theoretical_max == 255:
+            return "histogram/sensor_max_native_8"
+        if native_theoretical_max == 4294967295:
+            return "histogram/sensor_max_native_32u"
+        return "histogram/sensor_max_native_generic"
+
+    def _load_sensor_max_setting(self):
+        self.native_theoretical_max = self._theoretical_native_max_from_meta()
+        if self.native_theoretical_max:
+            key = self._settings_key_for_native_max(self.native_theoretical_max)
+            val = self.settings.value(key, None)
+            if val is not None:
+                try:
+                    self.sensor_native_max = float(val)
+                except Exception:
+                    self.sensor_native_max = None
+
+        self._recompute_effective_max01()
+
+    def _recompute_effective_max01(self):
+        if self.native_theoretical_max and self.sensor_native_max:
+            self.sensor_max01 = float(self.sensor_native_max) / float(self.native_theoretical_max)
+            self.sensor_max01 = float(np.clip(self.sensor_max01, 1e-6, 1.0))
+        else:
+            self.sensor_max01 = 1.0
+
+    def _prompt_sensor_max(self):
+        self.native_theoretical_max = self._theoretical_native_max_from_meta()
+
+        if self.native_theoretical_max:
+            key = self._settings_key_for_native_max(self.native_theoretical_max)
+            current = self.sensor_native_max or self.native_theoretical_max
+
+            val, ok = QInputDialog.getInt(
+                self,
+                self.tr("Sensor True Max (ADU)"),
+                self.tr("Enter your sensor's true saturation value in native ADU.\n"
+                "(Typical max for this file type is {0})\n\n"
+                "You can measure this by taking a deliberately overexposed frame\n"
+                "and reading its maximum pixel value.").format(self.native_theoretical_max),
+                int(current),
+                1,
+                int(self.native_theoretical_max)
+            )
+            if ok:
+                self.sensor_native_max = float(val)
+                self.settings.setValue(key, float(val))
+        else:
+            # float images / unknown depth: allow normalized max
+            val, ok = QInputDialog.getDouble(
+                self,
+                self.tr("Histogram Effective Max"),
+                self.tr("Enter effective maximum for clipping (normalized units)."),
+                float(self.sensor_max01),
+                1e-6,
+                1.0,
+                6
+            )
+            if ok:
+                self.sensor_max01 = float(val)
+                self.settings.setValue("histogram/sensor_max01_generic", float(val))
+
+        self._recompute_effective_max01()
+        self._update_stats()      # High Clipped row depends on sensor_max01
+        self._draw_histogram()
+
+    def _adjust_stats_width(self):
+        """Resize stats table so all columns are visible without a scrollbar."""
+        if not self.stats_table:
+            return
+
+        # Let Qt compute natural column widths
+        self.stats_table.resizeColumnsToContents()
+        self.stats_table.resizeRowsToContents()
+
+        vh = self.stats_table.verticalHeader()
+        frame = self.stats_table.frameWidth()
+
+        total_w = vh.width() + 2 * frame
+
+        for col in range(self.stats_table.columnCount()):
+            total_w += self.stats_table.columnWidth(col)
+
+        # Room for a possible vertical scrollbar
+        vbar = self.stats_table.verticalScrollBar()
+        if vbar is not None:
+            total_w += vbar.sizeHint().width()
+
+        # A tiny padding so text isn't tight
+        total_w += 6
+
+        self.stats_table.setMinimumWidth(total_w)
+
+
+    def _on_doc_destroyed(self, *args):
+        # Called when the owner/document goes away.
+        try:
+            # Avoid re-entrancy; schedule deletion safely.
+            self.deleteLater()
+        except RuntimeError:
+            pass
+
+    def closeEvent(self, event):
+        # Cleanly disconnect to avoid stray callbacks.
+        if getattr(self, "_doc_conn", False) and getattr(self, "doc", None) is not None:
+            try:
+                self.doc.destroyed.disconnect(self._on_doc_destroyed)
+            except (TypeError, RuntimeError):
+                pass
+            self._doc_conn = False
+
+        super().closeEvent(event)