go to [login to view URL] and download elastic vision. that is wat the software should look like you get....something close....i also attached another way to calculate the distance between the image histograms and the training model. the images will be too large to attach so if u can run the program u can try images from your side to build a new model
[login to view URL] HSV color histograms from the images
[login to view URL] into either 18*3*3 or any amonu of bins you want
[login to view URL] the pairwise distances between each image histogram by calculating the
[login to view URL] distances between the images histograms(and using norm as well)
then compute the pairwise distance for each pair of images so we generate a similarity matrix. this is a square matrix containing 1 as the diagonals because and image is completely similar to itself. you can also calculate a dissimilarity matrix instead and the diagonal will be 0 because every image are not dissimilar to theirselves.
[login to view URL] similarity matrix then serves as the input into the mds which is mdscale in matlab to generate a set of points in an euclidean space which best represent the distances of each image from the other meaning similar images are displayed very close to each while very dissimialr images are displayed far way. this images are then projected on the screen in thumbnails and on the locations which have been denoted by these points from the mds.
I have done more than 10 projects in matlab in image processing field likes inpainting, cbir, face Face Recognisation. this will be easy project for me.