Vision model based perceptual post filtering of JPEG2000 coded colour images

Abstract

This paper presents a perceptual post filtering coder for digital colour images in YCrCb colour space. The approach builds on our earlier perceptual coder (PC) and exploits intra-band and inter-orientation masking properties of Human Visual System (HVS) to identify, estimate and recover the amount of perceived visual information loss due to compression. The proposed technique applies to our earlier perceptual coder (PC) which retains most of the embedded Block Coding with Optimized Truncation (EBCOT) features and is bit-stream compliant to the JPEG2000 standard. We use PC coder to compress images with some information loss and hence loss of quality. The images are then reconstructed from the compressed bit-stream with our proposed post filtering coding technique that attempts to recover the perceived loss of visual information with a HVS model. The simulation results have shown that our proposed perceptual post filtering coder achieves comparable or superior visual performance over that of our PC, and that of JPEG2000 verification model 8.0 coder with both MSE and visual masking.