Hill-Climbing Color Image Segmentation

color image segmentation using hill-climbing algorithm and kmeans in CIE Lab.

Description

 Image is first converted to CIE lab color space and the 3d color histogram is generated. Hill-climbing algorithm is used to find the local maximums in lab histogram, which is used to automatically decide the cluster number K as well as initial seeds for K-means.

• Required Products: Statistics Toolbox
• MATLAB release: MATLAB 7.4 (R2007a)

[Anton Semechko]

very neat algorithm, does not require any explicit assumptions however performance still depends on one input parameter, namely the number of histogram bins (since the number of local maxima is likely be larger at higher resolutions). overall the code is very easy to read and its a very good submission. thanks

ps - couldnt find the first article in your references. if it all possible could you please forward it to me. my e-mail is [e-mail]. thanks once again