Fill the outside of contours OpenCV

Question:

I am trying to color in black the outside region of a contours using openCV and python language.
Here is my code :

contours, hierarchy = cv2.findContours(copy.deepcopy(img_copy),cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
areas = [cv2.contourArea(c) for c in contours]
max_index = np.argmax(areas)
cnt=contours[max_index]
# how to fill of black the outside of the contours cnt please? `
Asked By: Bill

||

Answers:

Here’s how you can fill an image with black color outside of a set of contours:

import cv2
import numpy
img = cv2.imread("zebra.jpg")
stencil = numpy.zeros(img.shape).astype(img.dtype)
contours = [numpy.array([[100, 180], [200, 280], [200, 180]]), numpy.array([[280, 70], [12, 20], [80, 150]])]
color = [255, 255, 255]
cv2.fillPoly(stencil, contours, color)
result = cv2.bitwise_and(img, stencil)
cv2.imwrite("result.jpg", result)

enter image description here
enter image description here

UPD.: The code above exploits the fact that bitwise_and with 0-s produces 0-s, and won’t work for fill colors other than black. To fill with an arbitrary color:

import cv2
import numpy

img = cv2.imread("zebra.jpg")

fill_color = [127, 256, 32] # any BGR color value to fill with
mask_value = 255            # 1 channel white (can be any non-zero uint8 value)

# contours to fill outside of
contours = [ numpy.array([ [100, 180], [200, 280], [200, 180] ]), 
             numpy.array([ [280, 70], [12, 20], [80, 150]])
           ]

# our stencil - some `mask_value` contours on black (zeros) background, 
# the image has same height and width as `img`, but only 1 color channel
stencil  = numpy.zeros(img.shape[:-1]).astype(numpy.uint8)
cv2.fillPoly(stencil, contours, mask_value)

sel      = stencil != mask_value # select everything that is not mask_value
img[sel] = fill_color            # and fill it with fill_color

cv2.imwrite("result.jpg", img)

enter image description here

Can fill with another image as well, for example, using img[sel] = ~img[sel] instead of img[sel] = fill_color would fill it with the same inverted image outside of the contours:

enter image description here

Answered By: Headcrab