ROI found automatically Image processing

Hello,

I would like to crop ROI from number of images automatically. I can't use the coordinates as are varied between images. I'm working on the ultrasonographic images, which have different tones of white, grey and black color and the background is not homogenous. I want the code to be able to find roughly circular structures of defined value of the grey tone and area. The difficulty lies in the fact that background, which is not roughly circular might have similar values of pixels and can touch my structure, what causes that the recognised image is not circular. I think that something as averaging the value of neighbourhood pixels could be good.

I wonder how to do it in Matlab? I've attached 2 examples of the images.

Thank you in advance for help.

Answer 1

[KALYAN ACHARJYA]

Sir the job is image segmenattion. As I see the attached first image to get the white portion, I feel thresholding may be sufficient. For more accurate results, you may apply some sort of preprocessing (histogram equlaizations) before thresholding, if not have you tried with other popular segmentation approaches (othshu etc). In case of unsupervised segmentation approach, you have provides certain conditions, which mostly satisfy by ROI only. Its quite hard to get 100% percent, mostly in medical images, which compropises lots of other disturbance in the image, there for you have to tradeoff between complexity of the code and accuracy (Higher Weightage).

[Malgorzata Wieteska]

Thank you for your answer. I thought that I can avoid the image segmentation but this is impossible. I've managed to remove small noise using Matlab commend bwareafilt