Efficient wavelet packet basis selection in JPEG2000

Abstract

Efficient rate-distortion optimal wavelet packet basis selection for the coding framework of JPEG2000 has been developed and evaluated. JPEG2000 Part 2 allows the application of arbitrary decomposition structures (wavelet packet bases). However, only a subset of all possible wavelet packet bases is permitted in the JPEG2000 standard. As we have developed a rate-distortion optimal wavelet packet basis selection algorithm, we can determine the cost of these restrictions for the first time. A further focus is on the efficient implementation of wavelet packet basis selection. The best basis selection in a rate-distortion sense is computationally feasible, yet not inexpensive. Several more efficient wavelet packet selection algorithms have been developed and are compared to the actually best solution. Our evaluations employ state-of-the-art quality assessment tools supplementary to PSNR evaluations.