Fix and improve the interpolation logic in resize_image()

The original code had a mistake. It used 'lanczos' when the image got smaller (width > resized_width and height > resized_height) and 'area' when it stayed the same or got bigger. This was the wrong way. 'area' is better for big shrinking.

We changed it so 'area' is used only when the image shrinks a lot (to half its size or smaller). This is because 'area' works best for making images much smaller. For other cases, 'lanczos' is used instead.

Author: LexSong

library/utils.py

@@ -421,8 +421,11 @@ def resize_image(image: np.ndarray, width: int, height: int, resized_width: int,
     resized_height = int(resized_height)
 
     if resize_interpolation is None:
-        resize_interpolation = "lanczos" if width > resized_width and height > resized_height else "area"
-    
+        if width >= resized_width * 2 and height >= resized_height * 2:
+            resize_interpolation = "area"
+        else:
+            resize_interpolation = "lanczos"
+
     # we use PIL for lanczos (for backward compatibility) and box, cv2 for others
     use_pil = resize_interpolation in ["lanczos", "lanczos4", "box"]