In this work, we propose and investigate techniques developed in the contex
t of fractal image coding for generalizing block-matching algorithms as use
d in motion-compensated video codecs. By allowing a gray value adaptation (
similar to non-iterative fractal techniques) within the block-matching proc
ess, we obtain MC error residuals with significantly lower entropy values a
s compared to pure block-matching. This leads to a significantly improved r
ate/distortion performance for video sequences with illumination changes or
high motion content at a low computational cost.