made autocrop work if the image was cropped previously

docs/notebooks/FAQ.ipynb

@@ -34,14 +34,6 @@
     "    gc.collect()  # manually calling the garbage collector after each iteration\n",
     "```"
    ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "d46c2043-de87-4b96-9db8-410b243aeef4",
-   "metadata": {},
-   "outputs": [],
-   "source": []
   }
  ],
  "metadata": {

spatialproteomics/pl/plot.py

@@ -1426,5 +1426,6 @@ def autocrop(
         if channel is None:
             channel = self._obj.coords[Dims.CHANNELS].values.tolist()[0]
         img = self._obj.pp[channel].pp.downsample(downsample)[key].values.squeeze()
-        slices = _autocrop(img, downsample=downsample, padding=padding)
+        bounds = self._obj.pl._get_bounds()
+        slices = _autocrop(img, bounds=bounds, downsample=downsample, padding=padding)
         return self._obj.pp[slices[0], slices[1]]

spatialproteomics/pl/utils.py

@@ -310,19 +310,19 @@ def _set_up_subplots(num_plots: int = 1, ncols: int = 4, width: int = 4, height:
     return fig, axes
 
 
-def _autocrop(img: np.ndarray, padding: int = 50, downsample: int = 10):
+def _autocrop(img: np.ndarray, bounds=List, padding: int = 50, downsample: int = 10):
     """
     Crop an image based on the regions of interest so that the background around the tissue/TMA gets cropped away.
 
     Parameters:
         img (np.ndarray): The input image as a NumPy array.
+        bounds (list): The bounds for the cropping. Should be given as [xmin, xmax, ymin, ymax].
         padding (int, optional): The padding to be added around the cropped image in pixels. Defaults to 50.
         downsample (int, optional): The downsample factor to be used for the cropping. Defaults to 10.
 
     Returns:
         tuple: A tuple containing two slices representing the cropped image.
     """
-
     bw = closing(img > np.quantile(img, 0.8), square(20))
     label_image = label(bw)
     props = regionprops(label_image)
@@ -336,10 +336,22 @@ def _autocrop(img: np.ndarray, padding: int = 50, downsample: int = 10):
         region = props[max_idx]
         minr, minc, maxr, maxc = region.bbox
 
-    return slice(downsample * minc - padding, downsample * maxc + padding), slice(
+    slices = slice(downsample * minc - padding, downsample * maxc + padding), slice(
         downsample * minr - padding, downsample * maxr + padding
     )
 
+    # getting the minimum x and y bounds
+    min_bounds = [bounds[0], bounds[2]]
+    # adding the offset from the bounds
+    slices = [slice(s.start + min_bounds[i], s.stop + min_bounds[i]) for i, s in enumerate(slices)]
+    # if the slices are outside of the bounds, we need to adjust them
+    slices = [
+        slice(max(bounds[0], slices[0].start), min(bounds[1], slices[0].stop)),
+        slice(max(bounds[2], slices[1].start), min(bounds[3], slices[1].stop)),
+    ]
+
+    return slices
+
 
 def _compute_erosion(segmentation: np.ndarray, erosion_strength: int = 5) -> np.ndarray:
     # Binarize the segmentation array: values > 0 become 1, others become 0