MATLAB Answers

What type technique should be used for segmenting the reference image given below?

1 view (last 30 days)
Celik on 31 Jul 2011
hi all
i am directly starting my interest;
the aim is to extract the red and green cells in the reference image. i have a problem like this;
in the image given the link below, red and green cells have to be seperated, i have achieved to extract red cells. But, green cells located in the middle of the image are mixed with noise. So, i fail in extracting the green cells.
if you have any (promising, hopeful, progressive) idea about this type of issue, please write.

Answers (1)

Image Analyst
Image Analyst on 31 Jul 2011
Is it always bright in the middle? Maybe you can model the background as a quadratic or Gaussian and subtract it. Otherwise I'd threshold the green channel around 69, then clean up the small things with bwareaopen, then call imfill. Maybe do some marker controlled watershed segmentation to split apart large objects if any are joined. Then maybe do some shape filtering to keep just the roundish objects. You can also try some noise filters like median, bilateral, etc. at various places in the process.
  1 Comment
Celik on 1 Aug 2011
Firstly, thanks for answer.
-yes, among images in my current study, the middle of images or right-up position are a little brighter and mixed with noise; already the problem i cannot solve at this point is green cells in the mixed(noise, or whatever it is) area . Because of this, thresholding can not generate good result for green cells.
-I'm sorry i dont understand what 'modelling the background as a quadratic or Gaussian and subtracting it' means? Probably, unless i'm right, substracting comes from there, the background will be modelled as a quadratic or Gaussian, and then it will be substracted from original. But still there is a ? for me about modelling the background as a quadratic of Gaussian.
-it is possible thresholding the green channel at 69, then cleaning small things using morphological operations, but where does 69 come, manuel input argument is not an option for me. if you tell me the reason(of 69) i'll have learnt and tried to automize.
- i have thought marker controlled watershed segmentation for latter steps (thanks), besides i tried median, averaging (etc. Gaussian filtering the green band for smoothing) for reducing noise.

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!