在最终图像上应用轮廓遮罩

Question

While trying to segment a plant from its background, I came across a problem After creating a mask by hue value and using the close and open operator on it, I get into the following situation:

After this I wanted to remove the small bits in edges of the image, I did this by the following operation:

_, cont, heir = cv2.findContours(mask_final, cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_NONE)
contour_sizes = [cv2.contourArea(contour) for contour in cont]
for con, size in zip(cont, contour_sizes):
    if size > 5000:
        mask_final = cv2.drawContours(mask_final, [con], -1, (255, 255, 255), cv2.FILLED)

When this is applied, the specks have been removed, but when applied by:

final = cv2.bitwise_and(img_rgb,img_rgb, mask = mask_final)

I get the following result:

As can be seen, the mask is not correctly applied on the image, does anybody know why this is happening?

Answer 1

Instead of using mask_final in cv2.drawContours function, create a new mask of same shape as original image and do something like below:

mask = np.zeros(img.shape, np.uint8)

for con, size in zip(contours, contour_sizes):
    if size > 5000:
        mask = cv2.drawContours(mask, [con], -1, (255, 255, 255), cv2.FILLED)

final = cv2.bitwise_and(img_rgb, img_rgb, mask = cv2.cvtColor(mask, cv2.COLOR_BGR2GRAY))