How to mask or crop an image region after applying connected components

Question

I have an RGB image (NxMx3 ndarray) having various regions. After applying clustering and connected components ( skimage measure.label ), I got an NxM ndarray with labeled regions.

How can I use these labeled areas for cropping or masking pixels from the original image (NxMx3)? In the end I want to take only the background (with label=0) and to extract some average color out of it.

Answer 1

Assuming image is your source image labels is your label image, you can grab the pixels corresponding to label j with:

pixels_j = image[labels == j]

This will give you an (Nj, 3) array of those pixels. You can then recover the mean color with:

mean_color = np.mean(pixels_j, axis=0)

If you are going to do this for many images, though, you should probably use skimage.measure.regionprops .