In this paper we investigate the effect of fast nearest neighbor search method on acceleration of fractal image compression. First we follow the Saupe1 encoding step of fractal image compression that uses the multi-dimensional nearest neighbor search in a projected space. Then we investigate performance of a method for finding the nearest vector called partial distortion elimination. We also propose some supplementary accelerating concepts. For all inspected methods the impact of the tolerance criterion for mean square error of block matching is examined.
Improving Approximate Nearest Neighbor Search through Learned Adaptive Early Termination