Shape detection of blobs

Hi,

I have two different types of blobs:

1. Close to kidney shape but not that perfect
2. Close to a rectangle but with a dent or two and with filleted vertices

How can I differentiate between the two using Image processing

Answer 1

[Image Analyst]

What I'd probably do is to get the bounding box. Then make a mask that is the bounding box and get the XOR of that with the blob. This will get you the black "bays" (black regions between blob and it's bounding box) as blobs. Then look at the area, and possibly aspect ratio (bounding box) of those blobs to determine which are like slivers, like the rectangular blob, and which are rounded and large like the bay in the kidney.

See my Image Segmentation Tutorial if you need a start.

[RIshabh Golchha]

Hi,

Thanks for your response. I did try a similar way. However, there is cases when the shape aren't oriented properly. For example, the rectangles "edges" are neither horizontal nor vertical. Is there some way to correct this?

Also, how do I differentiate slivers from rectangular blob ,and the rounded and large shapes from kidney shapes blobs?

[Image Analyst]

Actually I didn't mean bounding box, I meant convex hull. Try bwconvhull()

[RIshabh Golchha]

if I use

ran1 = bwconvhull(ran,'objects') & ~ran; % ran is the original binary image

I get the first image. And then if I use the following to remove thin regions, I get pretty identical shapes. (Although here, the two smaller blobs vary in size but that is not always the case. Hence I need a different parameter to differentiate them)

ran2 = imopen(ran1,strel('disk',2));