dliate和erode实现
import cv2 as cv
import numpy as np
pie=cv.imread('d:/pic/pie.png')
kernel=np.ones((7,7),np.uint8)
dilate=cv.dilate(pie,kernel,iterations=10)
erosion=cv.erode(pie,kernel,iterations=10)
res=np.hstack((pie,dilate,erosion))
cv.imshow('res',res)
cv.waitKey(0)
cv.destroyAllWindows()
结果如下:
可以发现膨胀和腐蚀圆形已经变形,向着正方形发展了,原因是kernel的形状是正方形,所以算法在计算的时候会越来越像正方形,
解决办法:
将kernel的形状改成圆形;
import cv2 as cv
import numpy as np
import math
pie = cv.imread('0000.png')
kernel = np.ones((7, 7), np.uint8)
kernel_re = []
rows, cols = kernel.shape
for i in range(rows):
result = [0 if math.sqrt((i-3)**2+(j-3)**2) > 3 else 1 for j in range(cols)]
kernel_re.append(result)
kernel_re = np.array(kernel_re, np.uint8)
print(kernel_re, kernel)
print(type(kernel_re), type(kernel))
print(kernel_re.shape, kernel.shape)
dilate = cv.dilate(pie, kernel_re, iterations=10)
erosion = cv.erode(pie, kernel_re, iterations=10)
res = np.hstack((pie, dilate, erosion))
cv.imshow('res', res)
cv.waitKey(0)
cv.destroyAllWindows()
结果如下:
可以看到结果没有那么圆,这个是因为kernel的形状不是标准圆造成的,但是已经对之前的结果有了相对明显的改善;