[FIXED] np.where select pixels that red value bigger than green value in an BGR image

Issue

It's my first time using np.where to select pixels from a bgr image, i have no idea how to select pixels that r>g from image by using np.where, I tried to do that by using codes like this:

bgr = cv2.imread('im.jpg')
bgr = np.where(bgr[1]>bgr[2],np.full_like(bgr,[255,255,255]),bgr)
cv2.imshow('result',bgr)
cv2.waitKey(0)

but it seems didn't work. Can anybody help me?

Solution

You seem to want to make all pixels white where red exceeds green and retain the original colours elsewhere. If so, I find it clearer this way:

# Load image
im = cv2.imread('image.png')

# Make simple synonyms for red and green
R = im[...,2]
G = im[...,1]

# Make True/False Boolean of pixels where red exceeds green
MoreRedThanGreen = R > G

# Make the image white wherever it is more red than green
im[MoreRedThanGreen,:] = [255,255,255]

# Save result
cv2.imwrite('result.png', im)

That makes this start image:

into this:

---

Note: Read the colon (:) in this line as meaning "all the channels":

im[MoreRedThanGreen,:] = [255,255,255]

Answered By - Mark Setchell