Entropy improvement for fractal image coder

Abstract EN

Fractal Image Coder (FIC) makes use of the self-similarity inside a natural image to achieve high compression ratio and maintain good image quality.

In FIC, the most important factor affecting the compression ratio and the image quality is the quantization of the contrast scaling and brightness offset coefficients.

Most quantization methods treat the two coefficients independently and quantize them separately.

However, the two coefficients are highly correlated and scatter around a line.

In this paper, a joint coefficient quantization method is proposed that considers the two coefficients together and thereby achieves better compression ratio and image quality.

The proposed method is especially effective under parsimonious conditions.

For example, using only 3 bits each to represent the contrast and brightness coefficients of Lena, the proposed method yields quality of 27.04 dB, which is significantly better than 22.87 dB obtained from the traditional linear quantization method.