Wavelet packet image coding with consideration of the self-similarities by zerotree structure

Abstract

This paper describes an image compression method associating a wavelet packet and a zerotree coding scheme. The wavelet packet, which gives a generalized multiresolution analysis, has been applied so as to reduce the correlation of the image pixels. To encode the adaptive wavelet packet coefficients, we apply the zerotree structure, which can reduce the correlation across the decomposed component images by exploiting the self-similarity among the components. In this case, the decomposition criterion with regard to each decomposition level is computed using a simplified local zerotree structure. If a target bit rate or a target distortion is decided, we can obtain the most suitable decompositions using an operation with fewer computations. Simulation results show that the proposed method has a better coding efficiency than the wavelet octave decomposition case by about 1 dB, and the JPEG baseline case by about 3 dB in PSNR at 0.4 bpp using the standard image 'BARBARA'.