JPEG2000 coded images optimization using a content-dependent approach

This paper presents a method to optimize the visual quality of JPEG2000-compressed images using an adaptive approach to assign more bits to regions in which errors are more visible, maintaining the global bit-rate unchanged. The encoder partitions the image into rectangular blocks called "tiles" to perform the compression. The proposed approach is based on the assumption that, by computing simple energy measures, it is possible to discern the most perceptually significant regions of the image and to code corresponding tiles properly. The effectiveness of the proposed algorithm depends on the accuracy of this classification that, in turn, depends on the tile's size. Experimental results show that the proposed technique yields a perceptive quality improvement of the image with no significant overhead in terms of complexity of the standard JPEG2000 encoder.