前景提取中的蒙版值

Question

I am working on this code to perform foreground extraction but I don't understand the meaning of mask2 = np.where((mask==2 | (mask == 0), 0, 1).astype('uint8')) and img = img*mask2[:, :, np.newaxis] lines. Here is the code

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

img = cv2.imread('tut12img.jpg')
mask = np.zeros(img.shape[:2], np.uint8)
bgdModel = np.zeros((1,65), np.float64)
fgdModel = np.zeros((1,65), np.float64)

rect = (161, 79, 150, 150)

cv2.grabCut(img, mask, rect, bgdModel, fgdModel, 5, cv2.GC_INIT_WITH_RECT)
mask2 = np.where((mask==2)|(mask==0),0,1).astype('uint8')
img = img*mask2[:, :, np.newaxis]
plt.imshow(img)
plt.colorbar()
plt.show()

Why musk is zero when it is 2 or 0? What is special about it? what kind of matrix is musk2? Thanks!

code and photo comes from https://pythonprogramming.net/grabcut-foreground-extraction-python-opencv-tutorial/?completed=/template-matching-python-opencv-tutorial/

Answer 1

mask2 is the foreground mask. so you only use foreground pixels which is done by setting every pixel 0 that is background, else 1

mask2 = np.where((mask==2)|(mask==0),0,1).astype('uint8')

Pixels are put into 4 classes.

• background (0)

• foreground (1)

• probably background (2)

• probably foreground (3)

Of course it would have been more obvious if the author of that code would have used the proper flags like cv2.GC_BGD instead of 0...

Also mask2 might not be the most informative name.

It is all in the OpenCV manual. All you have to do is google 1-2 words and read 1 paragraph...

http://docs.opencv.org/3.2.0/d7/d1b/group__imgproc__misc.html#ga909c1dda50efcbeaa3ce126be862b37f

mask Input/output 8-bit single-channel mask. The mask is initialized by the function when mode is set to GC_INIT_WITH_RECT. Its elements may have one of the cv::GrabCutClasses.

and

enum cv::GrabCutClasses

class of the pixel in GrabCut algorithm

GC_BGD an obvious background pixels

GC_FGD an obvious foreground (object) pixel

GC_PR_BGD a possible background pixel

GC_PR_FGD a possible foreground pixel