Quick Python implementation of FAST corner detector

skimage/feature/fast.py

@@ -0,0 +1,35 @@
+import numpy as np
+from ..util import img_as_float
+
+
+def corner_fast(image, n=9, threshold=0.15):
+
+    image = np.squeeze(image)
+    if image.ndim != 2:
+        raise ValueError("Only 2-D gray-scale images supported.")
+
+    image = img_as_float(image)
+    corner_mask = np.zeros(image.shape, dtype=bool)
+
+    test_pixels = np.asarray([[-3, 0], [-3, 1], [-2, 2], [-1, 3], [0, 3],
+                              [1, 3], [2, 2], [3, 1], [3, 0], [3, -1],
+                              [2, -2], [1, -3], [0, -3], [-1, -3],
+                              [-2, -2], [-1, -3]])
+
+    # TODO : Outsource to Cython
+    for i in range(3, image.shape[0] - 3):
+        for j in range(3, image.shape[1] - 3):
+            test_x = i + test_pixels[:, 0]
+            test_y = j + test_pixels[:, 1]
+            intensities = image[test_x, test_y]
+            low = intensities < image[i, j] - threshold
+            high = intensities > image[i, j] + threshold
+            low = np.concatenate(low, low)
+            high = np.concatenate(high, high)
+            # How to check if a sequence n * [True] exists in low/high ?
+            # if n * [True] in low or n * True in high:
+            #     corner_mask[i, j] = True
+    
+    corner_x, corner_y = np.where(corner_mask == True)
+    corners = np.dstack(corner_x, corner_y)
+    return corners