本文共 1313 字，大约阅读时间需要 4 分钟。

5.图像梯度处理

#图像梯度处理   Sobel算子  Schaar算子  Laplacian算子#加载OpenCVimport cv2import numpy as npdef imgShow(title,imgData):    cv2.imshow(title,imgData)    cv2.waitKey()    cv2.destroyAllWindows()img = cv2.imread(r"C:\Users\Administrator\Pictures\Camera Roll\lenna.jpg",cv2.IMREAD_GRAYSCALE)#Sobel算子求GX轴的边缘 cv2.Sobel(imgData,cv2.CV_64F,GX,GY,ksize=3)          cv2.CV_64F 65-128保留负值img_x = cv2.Sobel(img,cv2.CV_64F,1,0,ksize=3)img_x = cv2.convertScaleAbs(img_x)  #取绝对值 cv2.convertScaleAbs(img_x) img_y = cv2.Sobel(img,cv2.CV_64F,0,1,ksize=3)img_y = cv2.convertScaleAbs(img_y)  #取绝对值 cv2.convertScaleAbs(imgData) #将Gx与Gy边缘检测图像融合img_result = cv2.addWeighted(img_x,0.5,img_y,0.5,0)result = np.hstack((img_x,img_y,img_result))# imgShow(r"result",result)#Scharr算子  可以显示更多的细节边缘  cv2.Scharr(imgData,cv2.CV_64F,GX,GY) img_sx = cv2.Scharr(img,cv2.CV_64F,0,1)img_sx = cv2.convertScaleAbs(img_sx)img_sy = cv2.Scharr(img,cv2.CV_64F,1,0)img_sy = cv2.convertScaleAbs(img_sy)img_sresult = cv2.addWeighted(img_sx,0.5,img_sy,0.5,0)sresult = np.hstack((img_sx,img_sy,img_sresult))# imgShow(r"result",sresult)#Laplacian算子  对噪声点过于敏感  cv2.Laplacian(imgData,cv2.CV_64F)img_lap = cv2.Laplacian(img,cv2.CV_64F)img_lap = cv2.convertScaleAbs(img_lap)# imgShow(r"result",img_lap)ret = np.hstack((img_result,img_sresult,img_lap))imgShow(r"compare",ret)