How to extract the features from the region of interest in an image?

I am working on a medical images dataset. The problem I am facing is that the region of interest is present in different parts in all images. What I want is to extract the features from an image only from that particular region of interest. I do not want to do the cropping as it will degrade the quality of the ROI. I want to first tell the computer that it is the particular area (but this measurement of area must be same for all the images) from which I want to extract the features and with the help of a suitable feature descriptor I will extract the features?Would sliding window work? If yes then how? Please help me with this. Thanks in advance!!!!

[KALYAN ACHARJYA]

"I want to first tell the computer that it is the particular area"

Go for effective segmentation, the approach followed based on input image. Then your first step may be segmenation, then only feature extraction on segmented part, right?

[Image Analyst]

It totally depends on what your images look like. Please read this link then post two images that show the ROI in different locations and specify exactly what this thing is that you're looking for (it may not be obvious to us).

[shivasmic]

First of all thank you guys for reverting back!!!!

For example: In these images the tumor can be an ROI but as you can see in the upper image the tumor is present at the top while in the lower image the tumor is present in the bottom of the image. So I want to tell the computer that it is the ROI and from that ROI i want to extract the features. Now you might say that segmentation might work here but the thing is segmentation will convert them into a binary images and hence I would not be able to extract texture feature and second thing is that my algorithm does not work well for all the images in the data set since it is a large dataset. Thats why I have decided to tell the computer explicitely that look this is the ROI and for that I can fix the dimensions of the ROI for example 500X500, etc.

So when the program executes it asks me which is the ROI and then I select the particular region of a particular size and then I can extract features.

Thanks in advance!!!!

[shivasmic]

Hii, just got to know that I can use imfreehand() to draw the roi on the image that's what I wanted to tell the computer but now the challenge is how to extract the features just from that particular boundary from the image. Please help if you know something aboout that. Thanks in advance!!!

[KALYAN ACHARJYA]

Hii, just got to know that I can use imfreehand() to draw the roi on the image that's what I wanted to tell the computer

That means you manually use the imfreehand() to segment all images invividually?

[shivasmic]

Yes, but now I need to perform the feature extraction in that particular region only. Can you help me with that??