Running feature extraction on region within a boundary

Question

The image below shows a cow where the boundary has been detected using a combination of thresholding and subtracting a background from a 3D depth image.

My goal is to perform feature extraction on the area INSDIE the boundary. I have read the other questions and have struggled to implement the steps refereed to in similar questions. I do not want to extract the area in the boundary, I simply want to use it for feature extraction.

Please could someone offer a solution that is perhaps simpler? For example, is there a way to give the extractSURFFeatures the boundary coordinates from which to work within?

Below is my boundary code which recieves my processed thresholded image (BW1).

figure(1);
  imshow(ImageCell_int{i-269});
  %title('Outlines, from bwboundaries()'); axis square;
  hold on;
  boundaries = bwboundaries(BW1);   
  numberOfBoundaries = size(boundaries);
  for k = 1 : numberOfBoundaries
     thisBoundary = boundaries{k};
     plot(thisBoundary(:,2), thisBoundary(:,1), 'g', 'LineWidth', 2);
  end
  hold off;

I would be extremely grateful for any assistance on this.

Answer 1

Great, now I see the cow! :)

You cannot specify an irregularly-shaped region of interest for the detectSURFFeatures function. However, you can detect the features in the whole image, and then create a binary mask of the region of interest, and use it to exclude keypoints, which are outside it.

Edit: If your boundary is represented as a polygon, you can use roipoly function to create a binary mask from it.

Having said that, features that are outside your object's boundary can actually be useful, because they capture information about the shape of the object.

Also, what is your final goal? If you want to recognize individual cows, then local features may not be the best approach. You may do better with a global HOG descriptor (extractHOGFeatures) or with a color histogram, or both.