After spending some time going through ideas for possible image comparison algorithms, I thought of an idea that could compare parts of images to other parts of other images we would have in an image bank. This idea uses a concept of breaking an images into different shapes. These shapes could then be compared with each other in order to find similar shapes. These similar shapes would be inserted into the place of the shape of the original shape.

Attaining Shapes

Using straight line detection (hough line), it’ll be possible to get an outline of a part of the image. As this outline is a collection of connected straight lines, it will create a shape with n vertices and n-1 edges. This is then possible to take the angles of these shapes at the vertices and then compare them with angles of other shapes in different images. This is assuming that the images we have in our image bank have already been segmented into these shape objects.

If we had a shape match, then it would be possible to that region of the other image and place it into the original image. This new segment would be blended into the original image to try and give the appearance that the new segement is part of the original image.