add normalize_img function

Author: lucamarini22

pyraws/utils/img_processing_utils.py

@@ -34,3 +34,31 @@ def is_2D_image(img: np.ndarray) -> bool:
         bool: True if an image has just two dimensions. False otherwise.
     """
     return img.shape == 2
+
+
+def normalize_img(
+    img: np.ndarray,
+    min: float,
+    max: float,
+    a: float = 0,
+    b: float = 1,
+    tol: float = 1e-6,
+) -> np.ndarray:
+    """Normalizes an image from range [min, max] to range [a, b].
+
+    Args:
+        img (np.ndarray): image to normalize.
+        min (float): minimum of the previous range.
+        max (float): maximum of the previous range.
+        a (float, optional): minimum of the new range. Defaults to 0.
+        b (float, optional): maximum of the new range. Defaults to 1.
+        tol (float, optional): tolerance. Defaults to 1e-6.
+
+    Returns:
+        np.ndarray: normalized image.
+    """
+    assert max != min
+    img_norm = ((b - a) * ((img - min) / (max - min))) + a
+    assert np.count_nonzero(img_norm < a - tol) <= 0
+    assert np.count_nonzero(img_norm > b + tol) <= 0
+    return img_norm

pyraws/utils/keypoint_utils.py

@@ -1,6 +1,7 @@
 from pyraws.utils.img_processing_utils import (
     is_2D_image,
     three_channel_to_grayscale_img,
+    normalize_img,
 )
 import numpy as np
 import cv2
@@ -37,8 +38,8 @@ def get_matched_keypoints_sift(
 
     sift = cv2.SIFT_create()
 
-    img_1 = (img_1 - np.min(img_1)) * max_val_grayscale / np.max(img_1)
-    img_2 = (img_2 - np.min(img_2)) * max_val_grayscale / np.max(img_2)
+    img_1 = normalize_img(img_1, np.min(img_1), np.max(img_1), 0, max_val_grayscale)
+    img_2 = normalize_img(img_2, np.min(img_2), np.max(img_2), 0, max_val_grayscale)
     img_1 = img_1.astype(np.uint8)
     img_2 = img_2.astype(np.uint8)
     # Find keypoints and descriptors.