Wavelet packet basis selection for JPEG2000 based on EBCOT modeling

Traditional Discrete Wavelet Transform (DWT) image compression fails to reserve texture information. JPEG2000 part 2 extends to allow arbitrary subband decomposition and different decomposition direction. In this paper, we aim to provide a suitable wavelet packet basis selection for JPEG2000 by the improved cost function. Various cost functions used to search for the optimal wavelet basis are discussed and compared, and the estimated coding cost (ECC) is chosen as our criterion to achieve both accuracy and simplified computation. Then the entropy coding scheme of JPEG2000 is analyzed in detail and modeled to get our cost function. Experimental results show that compared with other cost function, the proposed cost function with ECC is more efficient in terms of PSNR (Peak Signal to Noise Rate). The proposed adaptive wavelet packet coding can provides higher visual quality than original JPEG2000.