当 Canny 图像如此清晰时，为什么使用 findContours() 找到的轮廓都被分解成碎片？

Question

左边是我的Canny Image。右边是我的原始图像，轮廓用不同颜色标记。

这让我很烦恼...为什么当 Canny Image 如此清晰时，cv2.findContours() 函数仍然没有给我一个 包裹大矩形的单个封闭轮廓 ??

原始输入图像：

我的代码：

# [get colored raw image]
rawIm = cv2.imread("sample1.jpg");

# [gray]
print("--------- Image Modification: Gray and Invert --------")
grayIm = cv2.cvtColor(rawIm, cv2.COLOR_BGR2GRAY);
grayIm = inverte(grayIm) # invert the black to white

# [blurring]
print("--------- Image Modification: Blur --------")
blurredIm = cv2.medianBlur(grayIm, 5);

# [dilate]
print("--------- Image Modification: Dilated --------")
kernel = np.ones((7,7), np.uint8)
dilatedIm = cv2.dilate(blurredIm, kernel, iterations=1)

# [get edges]
cannyIm = cv2.Canny(dilatedIm, lowThreshold, lowThreshold*cannyThresholdRatio)

# [FIND CONTOURS]
cnts, hier = cv2.findContours(cannyIm.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)

# [DRAW CONTOURS]
for cnt in cnts:
        color = (random.randint(0,255), random.randint(0,255), random.randint(0,255))
        cv2.drawContours(rawIm, [cnt], -1, color , 6)

cv2.imshow("CANNY", resize(cannyIm, percentageShrink))
cv2.imshow("CONTOURS", resize(rawIm, percentageShrink))

是这个问题吗？

cv2.findContours(cannyIm.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)

除了“cv2.RETR_LIST”和“cv2.CHAIN_APPROX_SIMPLE”之外，我必须使用不同的参数吗？

Answer 1

好的。扩张有效。在 Canny 过程之后添加 cannyIm = cv2.dilate(cannyIm, kernel, iterations=1) 就成功了。

完整代码

# [get colored raw image]
rawIm = cv2.imread("sample1.jpg");

# [gray]
print("--------- Image Modification: Gray and Invert --------")
grayIm = cv2.cvtColor(rawIm, cv2.COLOR_BGR2GRAY);
grayIm = inverte(grayIm) # invert the black to white

# [blurring]
print("--------- Image Modification: Blur --------")
blurredIm = cv2.medianBlur(grayIm, 5);

# [dilate]
print("--------- Image Modification: Dilated --------")
kernel = np.ones((7,7), np.uint8)
dilatedIm = cv2.dilate(blurredIm, kernel, iterations=1)

# [get edges]
cannyIm = cv2.Canny(dilatedIm, lowThreshold, lowThreshold*cannyThresholdRatio)
cannyIm = cv2.dilate(cannyIm, kernel, iterations=1)

# [FIND CONTOURS]
cnts, hier = cv2.findContours(cannyIm.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)

# [DRAW CONTOURS]
for cnt in cnts:
        color = (random.randint(0,255), random.randint(0,255), random.randint(0,255))
        cv2.drawContours(rawIm, [cnt], -1, color , 6)

cv2.imshow("CANNY", resize(cannyIm, percentageShrink))
cv2.imshow("CONTOURS", resize(rawIm, percentageShrink))

谢谢@fmw42

当 Canny 图像如此清晰时，为什么使用 findContours() 找到的轮廓都被分解成碎片？

Why contours found using findContours() are all broken up into pieces when the Canny Image is so clear?

python

opencv