Fractal image compression with adjacent block parameter estimation

Fractal image coding has been successfully applied to encode digital images at low bit rates. The usual coding technique is to partition a given image into a number of blocks (range blocks). Each block of the partition is expressed as the contractive transformation of another part of the image. However, such scheme does not take into account of the local smoothness in most images. We propose a novel fractal image coding scheme which estimates the parameters of the current block from those of the adjacent blocks. Each block is examined based on a criterion called the minimum edge difference (MED). If the MED criterion is fulfilled, a joint optimization of the adjacent and current block parameters or a predictive coding of the scaling and offset parameters of the current blocks is performed. Experiments show that a reduction of about 20% in the bit rates can be achieved with nearly no loss in the PSNR.