影象處理DOG 演算法,python結合cv2實現
阿新 • • 發佈:2019-02-18
DoG (Difference of Gaussian)是灰度影象增強和角點檢測的方法
#coding=utf-8
import cv2
import numpy as np
def getExtrema(A, B, C, thresh):
height,width= A.shape
resu = np.ones((height, width), A.dtype) * 100
for row in range(1, height-1):
for col in range(1, width-1):
center = B[row, col]
if center < thresh:
continue
B[row, col] = B[row, col - 1]
minValue = np.vstack([A[row-1:row+2, col-1:col+2], B[row-1:row+2, col-1:col+2],C[row-1:row+2, col-1:col+2]]).min()
maxValue = np.vstack([A[row - 1:row + 2, col - 1:col + 2], B[row - 1:row + 2 , col - 1:col + 2],
C[row - 1:row + 2, col - 1:col + 2]]).max()
if center < minValue:
resu[row, col] = 0
if center > maxValue:
resu[row, col] = 255
B[row, col] = center
return resu
def addPoint (image, image_point):
height, width, dvim = image.shape
for row in range(0, height):
for col in range(0, width):
if image_point[row, col] == 255:
cv2.circle(image, (row, col), 5, thickness=1, color=[0,0,255])
elif image_point[row, col] == 0:
cv2.circle(image, (row, col), 5, thickness=1, color=[0,255,0])
if __name__ == "__main__":
image = cv2.imread('lena.jpg')
r,g,b = cv2.split(image)
image_gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
image_gray_blur1 = cv2.GaussianBlur(image_gray, (3, 3), 0.3)
image_gray_blur2 = cv2.GaussianBlur(image_gray, (3, 3), 0.4)
image_gray_blur3 = cv2.GaussianBlur(image_gray, (3, 3), 0.5)
image_gray_blur4 = cv2.GaussianBlur(image_gray, (3, 3), 0.6)
image_gray_blur5 = cv2.GaussianBlur(image_gray, (3, 3), 0.7)
image_gray_blur6 = cv2.GaussianBlur(image_gray, (3, 3), 0.8)
image_gray_dog1 = image_gray_blur2 - image_gray_blur1
image_gray_dog2 = image_gray_blur4 - image_gray_blur3
image_gray_dog3 = image_gray_blur6 - image_gray_blur5
image_point = getExtrema(image_gray_dog1, image_gray_dog2, image_gray_dog3, 2)
#反過來的gbr
cv2.namedWindow("image_DOG", flags= cv2.WINDOW_NORMAL)
cv2.moveWindow("image_DOG", 300, 200)
addPoint(image, image_point)
cv2.imshow("image", cv2.imread("./lena.jpg"))
cv2.imshow("image_gray", image_gray)
cv2.imshow("image_gray_blur1", image_gray_blur1)
cv2.imshow("image_gray_blur2", image_gray_blur2)
cv2.imshow("image_gray_blur3", image_gray_blur3)
cv2.imshow("image_gray_blur4", image_gray_blur4)
cv2.imshow("image_gray_blur5", image_gray_blur5)
cv2.imshow("image_gray_blur6", image_gray_blur6)
cv2.imshow("image_gray_dog1", image_gray_dog1)
cv2.imshow("image_gray_dog2", image_gray_dog2)
cv2.imshow("image_gray_dog3", image_gray_dog3)
cv2.imshow("image_DOG", image)
cv2.imwrite("image_gray.jpg", image_gray,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_blur1.jpg", image_gray_blur1,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_blur2.jpg", image_gray_blur2,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_blur3.jpg", image_gray_blur3,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_blur4.jpg", image_gray_blur4,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_blur5.jpg", image_gray_blur5,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_blur6.jpg", image_gray_blur6,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_dog1.jpg", image_gray_dog1,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_dog2.jpg", image_gray_dog2,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_gray_dog3.jpg", image_gray_dog3,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
cv2.imwrite("image_DOG.jpg", image,[int(cv2.IMWRITE_JPEG_QUALITY), 100] )
cv2.waitKey(0)
cv2.destroyAllWindows()
ean.jpg
image_gray.jpg
image_gray_blur1.jpg
image_gray_blur2.jpg
image_gray_blur3.jpg
image_gray_blur4.jpg
image_gray_blur5.jpg
image_gray_blur6.jpg
image_gray_dog1.jpg
image_gray_dog2.jpg
image_gray_dog3.jpg
image_DOG.jpg