A Technique for Faster Fractal Image Coding

Abstract

This paper presents strategies for improving the Compression Ratio and speed of fractal image coding techniques. The technique of fractal image coding is computationally expensive, as it explores the piecewise self-similarity of the image. A codebook of m domain blocks (portions of the image) is searched exhaustively to find a best match for another portion of the image, called the range block. To optimize on the search speed we classify the codebook of m domain blocks using the values of mean and variance of the blocks. Each class contains a small number of domain blocks, hence for n range blocks a search complexity of nearly O (n) time is achieved. The proposed triangulation scheme increases the compression ratio while, maintaining the quality of reconstruction. Contrast and brightness transformation coefficients are used to adjust the intensity values in the domain to those in the range. The brightness transformation coefficients vary smoothly across an image and are subjected to dct-based coding. On an average the triangulation scheme offers a compression ratio of 60: 1 and a 30 dB PSNR for reconstructed images.

Indexing terms:

• Compression
• Fractal
• Triangulation