nec c

comic_panel_extractor/border_panel_extractor.py

@@ -12,7 +12,7 @@ import cv2
 
 from .config import Config
 from .image_processor import ImageProcessor
-from .utils import remove_duplicate_boxes
+from .utils import remove_duplicate_boxes, count_panels_inside
 
 class BorderPanelExtractor:
     """
@@ -241,6 +241,25 @@ class BorderPanelExtractor:
 
         return adjusted_boxes
 
+    def remove_swallow_boxes(self, boxes):
+        filtered_boxes = []
+
+        for i, (x1, y1, x2, y2) in enumerate(boxes):
+            current_box = (x1, y1, x2, y2)
+            # Count how many other boxes are fully inside this one
+            inside_count = count_panels_inside(current_box, [b for j, b in enumerate(boxes) if j != i])
+
+            # Skip this box if it fully contains at least one other box (i.e., it's swallowing)
+            if inside_count >= 1:
+                continue
+
+            # Keep boxes that don't swallow others
+            filtered_boxes.append(current_box)
+
+        print(f"✅ Found {abs(len(filtered_boxes) - len(boxes))} swallowed boxes")
+        return filtered_boxes
+
+
     def create_image_with_panels_removed(
         self,
         original_image: np.ndarray,
@@ -284,6 +303,8 @@ class BorderPanelExtractor:
 
         accepted_boxes = remove_duplicate_boxes(accepted_boxes)
 
+        accepted_boxes = self.remove_swallow_boxes(accepted_boxes)
+
         all_paths = self._save_panel(original_image, accepted_boxes)
 
         output_path = self.draw_black(original_image, accepted_boxes)

comic_panel_extractor/image_processor.py

@@ -9,6 +9,10 @@ from skimage.measure import label
 from skimage import measure
 from tqdm import tqdm
 
+from PIL import Image
+import numpy as np
+from sklearn.cluster import KMeans
+
 class ImageProcessor:
     """Handles image preprocessing operations."""
     
@@ -81,6 +85,41 @@ class ImageProcessor:
         # Save result
         return inverted, black_pixels > white_pixels
 
+    def group_colors(self, processed_image_path, num_clusters: int = 5, file_name="group_colors.jpg", output_folder=None) -> Image.Image:
+        """
+        Groups similar colors in an image using KMeans clustering.
+
+        Args:
+            processed_image_path (str): Path to the image to be color-grouped.
+            num_clusters (int): Number of color clusters to form.
+            file_name (str): Name of the output image file.
+            output_folder (str): Optional output directory.
+
+        Returns:
+            str: Path to the saved grouped-color image.
+        """
+        output_folder = output_folder or self.config.output_folder
+        # Load image
+        image = Image.open(processed_image_path).convert("RGB")
+        np_image = np.array(image)
+        h, w = np_image.shape[:2]
+        pixels = np_image.reshape(-1, 3)
+
+        # Run KMeans
+        kmeans = KMeans(n_clusters=num_clusters, random_state=42, n_init='auto')
+        labels = kmeans.fit_predict(pixels)
+        centers = kmeans.cluster_centers_.astype(np.uint8)
+
+        # Replace pixels with their cluster center color
+        clustered_pixels = centers[labels].reshape(h, w, 3)
+
+        # Save using OpenCV (convert RGB to BGR)
+        output_path = self.get_output_path(output_folder, file_name)
+        clustered_bgr = clustered_pixels[:, :, ::-1]
+        cv2.imwrite(output_path, clustered_bgr)
+
+        return str(output_path)
+
     def thin_image_borders(self, processed_image_path: str, file_name="thin_border.jpg", output_folder=None) -> str:
         """
         Clean dilated image by thinning thick borders and removing hanging clusters.

comic_panel_extractor/panel_extractor.py

@@ -6,7 +6,7 @@ import cv2
 from dataclasses import dataclass
 import os
 import re
-from .utils import remove_duplicate_boxes
+from .utils import remove_duplicate_boxes, count_panels_inside
 
 @dataclass
 class PanelData:
@@ -248,15 +248,6 @@ class PanelExtractor:
                 coords.append(tuple(map(int, match.groups())))
         return coords
 
-    def count_panels_inside(self, target_box, other_boxes):
-        x1a, y1a, x2a, y2a = target_box
-        count = 0
-        for x1b, y1b, x2b, y2b in other_boxes:
-            if x1a <= x1b and y1a <= y1b and x2a >= x2b and y2a >= y2b:
-                count += 1
-        return count
-
-
     def _save_panels(self, panels: List[Tuple[int, int, int, int]], original: np.ndarray, width: int, height: int) -> Tuple[List[np.ndarray], List[PanelData], List[str]]:
         """Save panel images and return panel data."""
         visual_output = original.copy()
@@ -310,7 +301,7 @@ class PanelExtractor:
                 continue
 
             # 2. Skip if this panel contains ≥1 other panels
-            contained_count = self.count_panels_inside((x1, y1, x2, y2), already_saved_coords)
+            contained_count = count_panels_inside((x1, y1, x2, y2), already_saved_coords)
             if contained_count >= 1:
                 print(f"⚠️ Skipping panel #{idx} — contains {contained_count} other panels inside")
                 continue

comic_panel_extractor/utils.py

@@ -58,4 +58,12 @@ def remove_duplicate_boxes(boxes, compare_single=None, iou_threshold=0.7):
             unique_boxes.append(box)
 
     print(f"✅ Found {abs(len(unique_boxes) - len(boxes))} duplicates")
-    return unique_boxes
+    return unique_boxes
+
+def count_panels_inside(target_box, other_boxes):
+    x1a, y1a, x2a, y2a = target_box
+    count = 0
+    for x1b, y1b, x2b, y2b in other_boxes:
+        if x1a <= x1b and y1a <= y1b and x2a >= x2b and y2a >= y2b:
+            count += 1
+    return count