Approximate treatment for calculation of the rate-distortion slope in EBCOT

Some novel approaches are presented aiming at reducing computation and memory for the EBCOT algorithm in JPEG2000. In the current structure of EBCOT, the algorithm for calculating the distortion reduction is still too complex for hardware-oriented design, and the memory space needed for the storage of large amounts of rate-distortion slopes is often unaffordable. In this paper, we propose a set of simplification schemes based on some approximate treatments by means of quantization median substitution and logarithm expression. These approaches can facilitate calculation of the rate-distortion slopes and reduce the memory by 50% (only 8 bits for each slope). Experimental results show that the quality of decoded images has only slightly decreased, which is negligible for most of the applications.