Image coding with fractal vector quantization

Abstract

In this paper, we address the time complexity problem associated with fractal image coding. In particular, we describe a new hybrid technique called Fractal Vector Quantization (FVQ), which takes advantage of the best qualities in fractal coding and vector quantization (VQ). In our proposed approach, VQ is used to construct a set of real world building blocks which can be used to approximate an arbitrary image. Fractal coding is then employed to fractalize the building blocks by finding an affine transformation for each block which best describes the block. The real world building blocks with their affine transformations are compiled in a fractal dictionary. To encode an image, FVQ approximates the image with a set of affine transformations from the precompiled fractal dictionary. The decoder uses a standard fractal decoding algorithm since the fractal dictionary is not required by the decoder.