Dear Muhammad, first you could use bwlabel command to label the connected components in your binary image to identify them. Then, you could use regionprops command to calculate properties such as area, perimeter, bounding box, centroid, and many more.

The command regionprops works directly on BW images by identifying the connected regions and extracting their properties. That is why I suggest to label the connected shapes (regions) to calculate their properties without confusion.

Oh, you should have the Image Processing Toolbox to use these commands. I am assuming you got it.

Cheers.

Dear Muhammad, first you could use bwlabel command to label the connected components in your binary image to identify them. Then, you could use regionprops command to calculate properties such as area, perimeter, bounding box, centroid, and many more.

The command regionprops works directly on BW images by identifying the connected regions and extracting their properties. That is why I suggest to label the connected shapes (regions) to calculate their properties without confusion.

Oh, you should have the Image Processing Toolbox to use these commands. I am assuming you got it.

Cheers.

You may follow these below links.....

https://in.mathworks.com/help/images/ref/bwperim.html?requestedDomain=www.mathworks.com

http://stackoverflow.com/questions/14572973/get-length-of-irregular-object-in-a-bw-or-rgb-picture-and-draw-it-into-picture-f

bwlabel with regionprops will work as suggested above.

if u want to write your own code, u will need to make clusters of 8 connected components and label the clusters which correspond to different objects.

then, u can just count the number of '1' pixels in each cluster to get the area.

to get the perimeter, u will need to calculate the number of boundary pixels in each cluster (boundary pixels will have a '0' valued neighbor in their 8 connected neighborhood, unlike interior pixels)

Thanks for your very useful guidance. It worked well. However, for some features, I get area = 1 with perimeter = 0. I don't understand this. What does this mean? How perimeter can be zero?

Dear Muhammad, it seems that some "regions" are single pixels, or even segments, lines, open curves, all of them BW noise maybe. If you have labeled them before, just filter your results by focusing on those in which you are interested. I can see in some of these "noisy" parts in your image. If this a BW version of your grayscale image, maybe you should do a thresholding first to bound better your regions, unless you need those parts.

First, you should exclude the small noisy blob using thresholding mechanism, subsequently you can apply bwlabel and regionprops. 

Hi Muhammed,

Not sure if this is still relevant, but I wanted to introduce you to MIPAR (http://MIPAR.us) which could greatly simplify your life and work. Since your image was already segmented, I simply opened it up, added a couple steps to removed small features and noise, and with a switch of a dropdown, labelled each feature and randomly colored them. I have attached the Recipe file to do this (which you load into the Image Processor after opening the image).

Then I used the Color By Measure tool to color each feature according to Convex Area/Area (a measure of irregularity), and Area. Many more measurements are available.

Finally, MIPAR could serve as an all-in-one, segmenting+labelling+measuring solution for you if you are interested. Contact me any time with questions.

(Side Note: MIPAR is written entirely in MATLAB - but is installed as a standalone executable. Not sure if there is the potential for collaboration since you mentioned you use MATLAB)

Cheers,

John

Thank you John for sharing your experience and information about  MIPAR. Certainly it will be  useful for future work.