On the performance of fractal compression with clustering

The paper investigates a technique to reduce the computational complexity of fractal image compression on gray-scale images. The technique uses a clustering process on image domain blocks with the clusters formed with the use of k-d trees and the fast pairwise nearest neighbor algorithm of Equitz (1984). Results indicate the method is effective for smaller domain block sizes and generally shows improvement in terms of picture peak signal-to-noise ratio (SNR) over the quadrant variance classification method.