A naive approach would be to extract "red" colors from the image:
I = imread('peppers.png');
R = I(:,:,1);
G = I(:,:,2);
B = I(:,:,3);
% Here I am naively defining "red" as the red layer being greater than
% some threshold and the blue and green layers being less than that threshold
thold = 100;
mask = R>thold & G<thold & B<thold;
R(~mask) = 255;
G(~mask) = 255;
B(~mask) = 255;
J = cat(3,R,G,B);
figure; montage(cat(4,I,J),'Size',[1 2]);
More advanced approaches exists (google "image segmentation"). For instance, see the MATLAB example for color-based segmentation using k-means clustering.
