OpenCV Python，去除阴影得到轮廓

Question

I need to get circles' contour from the image. But there is a shadow gradient.(top is bright, bottom is dark) So topside background's GL is same as bottom side circle's GL.

I also have the back side image without circle and try this code

import cv2
import matplotlib.pyplot as plt

src = './test.jpg'
img = cv2.imread(src, cv2.IMREAD_GRAYSCALE)
back_src = './back.jpg'
back = cv2.imread(back_src, cv2.IMREAD_GRAYSCALE)

diff_img = 255-cv2.absdiff(img, back)
cv2.normalize(diff_img, norm_img, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_8UC1)

th, threshed = cv2.threshold(norm_img, 245, 255, cv2.THRESH_BINARY)

In this image, if i set threshold 245, i can get perfect binary image. But in the real images(grayscale), they are noisy so it's hard to set each threshold.

I think there's a better way. please help me....

test.jpg back.jpg

Answer 1

You can use adaptive thresholding for this:

gray = cv2.imread('/path/to/your/image/circles_0001.jpeg', cv2.COLOR_BGR2GRAY)

th = cv2.adaptiveThreshold(gray,255, cv2.ADAPTIVE_THRESH_MEAN_C,cv2.THRESH_BINARY_INV,11,2)