0

I have a question regarding a paper I read the other day.

Suppose that we have a black and white image.

This image contains a very large area with white pixels (and it is the area that we are primarily interested in).

The problem is that there are some other areas (way smaller) around this big area that are also white, and we need these small areas removed.

They claim that they removed these areas by region based segmentation. They found the total amount of white pixels , and used a threshold, to remove every areas with less than 20% of the total white pixels.

My question is; How is this possible?

Maybe I didn't understood the paper correctly ?

Thanks in advance.

Mrswan
  • 1
  • 1
    Could you post the title of the paper? – David Nilosek Dec 04 '13 at 22:05
  • 3
    You can use `regionprops` with `area` (number of pixel of the regions) and `centroid` (coordinates) – marsei Dec 04 '13 at 22:08
  • 1
    Macduff is right, this is very easy with binary (black/white) images. Just use `regionprops` or `bwconncomp`. See [this question](http://stackoverflow.com/questions/12688524/connected-components-in-opencv) for the OpenCV equivalent. – chappjc Dec 04 '13 at 23:12

0 Answers0