I need some hints/tips regarding some project I'm doing. The project is about finding shapes in digital photos (JPEG file). Here's what I have to do: A digital photo is a collection of pixels each with a colour attribute. Each pixel apart from the edges of the photo has 4 neighbours (2 vertical pixels, 2 horizontal). The strength of an edge between neighbouring pixels in the graph model of the photo depends on how similar the colour attributes are (edge strength). The problem is to use this information to separate the main different coloured shapes in the photo. For example, it should be possible to pick out a red disk on a white background (Japanese flag). The first problem you will encounter, will be how to get the pixel colour attributes from the JPEG file. If you think you can solve this, then there are many choices for the next step (getting the shapes). One I am particularly interested in is using short random walks, moving from pixel to pixel randomly, but depending on the edge strength.
I'm thinking about using the PixelGrabber class in Java, but any other ideas are more than welcomed.
Thank you very much.