Parameter quantization in fractal image coding

The present paper considers quantization of the parameters in a block-based fractal coding algorithm. First, the coding principle is explained. In contrast to more commonly used fractal coder variants, our method yields uncorrelated parameters within each image block. We analyze the statistical properties of the parameters in question, and suggest scalar quantization techniques based on these properties, as well as on a simple mathematical analysis of the effects of quantization. Finally, the quantization strategies suggested are tried out on a number of real world images. The results show that the loss of quality relative to using unquantized parameters is quite low.<>