Vision model based perceptual coding of digital images

摘要

This paper presents a perceptual coder based on the Embedded Block Coding with Optimised Truncation (EBCOT) structure which visually outperforms JPEG2OOO VM 8.0 utilising Mean-Square-Error (MSE) criterion. Furthermore, the proposed perceptual coder shows a performance comparable to or better than CVIS, the EBCOT implementation with visual masking. This performance gain is attributed to more advanced vision modelling which includes intra-band and cross orientational masking.