filling contours with opencv python

Question

I have binary image with polylines created with:

cv2.polylines(binaryImage,contours,1, (255,255,255))

What I need now is effective method to fill all polylines. I haven't found such method in opencv, but maybe it exists. Alternatively, maybe I could implement algorithm to do the job (but fast one- I have HD ready pictures). Please share your thoughts..

Answer 1

I think what you are looking for is cv2.fillPoly, which fills the area bounded by one or more polygons. This is a simple snippet, I generate a contour of four points representing vertices of a square, then I fill the polygon with a white color.

import numpy as np
import cv2

contours = np.array( [ [50,50], [50,150], [150, 150], [150,50] ] )
img = np.zeros( (200,200) ) # create a single channel 200x200 pixel black image 
cv2.fillPoly(img, pts =[contours], color=(255,255,255))
cv2.imshow(" ", img)
cv2.waitKey()

Answer 2

Use cv2.drawContours() with thickness=cv2.FILLED:

cv2.drawContours(img, contours, -1, color=(255, 255, 255), thickness=cv2.FILLED)

Answer 3

You can use fillPoly or drawContours if your contour is closed. Pulling together @jabaldonedo and @ash-ketchum answers:

import cv2
import matplotlib.pyplot as plt
import numpy as np

# Lets first create a contour to use in example
cir = np.zeros((255,255))
cv2.circle(cir,(128,128),10,1)
res = cv2.findContours(cir.astype(np.uint8), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
contours = res[-2] # for cv2 v3 and v4+ compatibility

# An open circle; the points are in contours[0]
plt.figure()
plt.imshow(cir)

# Option 1: Using fillPoly
img_pl = np.zeros((255,255))
cv2.fillPoly(img_pl,pts=contours,color=(255,255,255))
plt.figure()
plt.imshow(img_pl)

# Option 2: Using drawContours
img_c = np.zeros((255,255))
cv2.drawContours(img_c, contours, contourIdx=-1, color=(255,255,255),thickness=-1)
plt.figure()
plt.imshow(img_c)

plt.show()

both img_pl and img_c contain a filled in circle from the points in contour[0]

For context, this was tested with python 3.6.2, OpenCV (cv2.version) 3.2.0, numpy 1.13.1 and matplotlib 2.0.2. I suspect it would work with anything cv2 3+ and python 3.5+. Per @elyas-karimi (and the OpenCV docs) findContours returns 3 values in 3.* and 2 in 4.* dropping the image return (and since 3.2 unmodified).