A filter Emboss Introduction:
Emboss filter used in edge detection and contour images, can effectively enhance the high frequency information (edges and contours) of the image, preserving the low frequency information (image content) of the image.
Emboss filter convolution ↑
Two experiments: Emboss filter implemented using filters enhance contours Emboss
1 import cv2 2 3 import numpy as np 4 5 # Gray scale 6 7 def BGR2GRAY(img): 8 9 b = img[:, :, 0].copy() 10 11 g = img[:, :, 1].copy() 12 13 r = img[:, :, 2].copy() 14 15 # Gray scale 16 17 out = 0.2126 * r + 0.7152 * g + 0.0722 * b 18 19 out = out.astype(np.uint8) 20 21 return out 22 23 # emboss filter 24 25 def emboss_filter(img, K_size=3): 26 27 H, W = img.shape 28 29 # zero padding 30 31 pad = K_size // 2 32 33 out = np.zeros((H + pad * 2, W + pad * 2), dtype=np.float) 34 35 out[pad: pad + H, pad: pad + W] = img.copy().astype(np.float) 36 37 tmp = out.copy() 38 39 # emboss kernel 40 41 K = [[-2., -1., 0.],[-1., 1., 1.], [0., 1., 2.]] 42 43 # filtering 44 45 for y in range(H): 46 47 for x in range(W): 48 49 out[pad + y, pad + x] = np.sum(K * (tmp[y: y + K_size, x: x + K_size])) 50 51 out = np.clip(out, 0, 255) 52 53 out = out[pad: pad + H, pad: pad + W].astype(np.uint8) 54 55 return out 56 57 # Read image 58 59 img = cv2.imread("../paojie.jpg").astype(np.float) 60 61 # BGR2GRAY 62 63 gray = BGR2GRAY(img) 64 65 # emboss filtering 66 67 out = emboss_filter(gray, K_size=3) 68 69 # Save result 70 71 cv2.imwrite("out.jpg", out) 72 73 cv2.imshow("result", out) 74 75 cv2.waitKey(0) 76 77 cv2.destroyAllWindows()
Three results:
Emboss filter after image effect ↑
After the picture is converted to grayscale ↑
Can be seen, Emboss filter can effectively enhance the contour of the image.
Four reference: