Added Harris Feature Detector to Digital Image Processing

digital_image_processing/feature_detectors/README.md

@@ -0,0 +1,20 @@
+# Computation of Harris Detector
+
+## Dependencies
+
+- opencv-python
+- Numpy
+- Scipy
+
+## Steps
+
+Given image $I$, $n\times n$ size Gaussian Kernel $G_{n\times n}$,
+
+1. Compute the gradients of the image, both horizontal and vertical directions. $X=(-1, 0, 1)\otimes I​$, $Y=(-1, 0, 1)^T \otimes I​$
+2. Compute the matrix $M$, where $A = G_{n\times n} \otimes X^2$, $B=G_{n\times n}\otimes Y^2$, $C=G_{n\times n}\otimes XY$
+3. Compute the response function $R​$, where $R=AB-C^2-k(A+B)​$
+4. Classify all points in $R​$.
+
+## Reference
+
+C. Harris and M. Stephens, “A Combined Corner and Edge Detector,” in Proceedings of Alvey Vision Conference 1988, Manchester, 1988, pp. 23.1-23.6.

digital_image_processing/feature_detectors/harris.py

@@ -0,0 +1,35 @@
+"""
+Implementation of Harris Detector
+
+Reference:
+C. Harris and M. Stephens, “A Combined Corner and Edge Detector,”
+in Proceedings of Alvey Vision Conference 1988, Manchester, 1988, pp. 23.1-23.6.
+"""
+
+import cv2 as cv
+import numpy as np
+from scipy.signal import convolve2d
+
+# Read original image and get gaussian kernel
+img = cv.imread('example.png')
+img_gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
+
+G = cv.getGaussianKernel(3, 1)
+
+# Begin Harris Detector Computation
+
+I = img_gray.astype('float32')
+X = convolve2d(I, [[-1, 0, 1]], mode='same')
+Y = convolve2d(I, [[-1], [0], [1]], mode='same')
+A = convolve2d(X*X, G, mode='same')
+B = convolve2d(Y*Y, G, mode='same')
+C = convolve2d(X*Y, G, mode='same')
+R = A*B - C*C - 0.04 * (A + B)
+
+# Finished Harris Detector Computation
+
+# Classify all points in R and show corners in image
+maxima = np.max(R)
+img[R > maxima*0.1] = [255, 0, 0]
+cv.imshow('harris_detector', img)
+cv.waitKey(0)