Study On Wavelet Coding Algorithm Based On Texture Analysis For Core Image

Now the core images are mostly compressed by the DCT-based JPEG algorithm, which will cause the fuzzy block-effects in the area of texture or edge regions in low bit rate situation. In this article, from the point of view for texture analysis, the method of wavelet packet decomposition and encoding algorithms for core image are studied. The following three aspects are included in this article:(1) The concept of edge active measure (EAM) is extended to the subband edge active measure (SEAM), which is used as the measurement of texture and cost-function for best basis selection of wavelet packet decomposition. The core image is decomposed by SEAM-based wavelet packet in order to reduce the SEAM in the high-frequency subbands, and enhance the similarity of coefficients within a subband, and thereby provide advantages to wavelet coding for core images later.(2) Modified SPECK algorithm for SEAM-based wavelet packet image coding is proposed. The method of S-octave band partitioning is more appropriate to fundamental distribution characteristics of coefficients in wavelet packet decomposition.The results show that the proposed algorithm provides gain in PSNR over base-line SPECK coder for core images, and better visual quality for textured regions in the core images.(3) Modified JPEG2000 algorithm for SEAM-based wavelet packet image coding is proposed. Combined with the regulations and restrictions of code block size in EBCOT algorithm, the parameter of code block size is set dynamically, which is more suited to the similarity of coefficients in the subbands of high frequency. The results show that the proposed algorithm provides gain in PSNR over JPEG2000 standard for core images, and better visual quality for textured regions in the core image .