CVSL Logo
FrancaisHome
AboutPeopleResearchPublicationsEventsProfile
About
Publications

 

 

 

REPARTI

MIVIM

A masking model for scalar non-fixed deadzone quantization in lossy JPEG2000 compression


Thierry Eude and Christophe Charrier


Abstract - JPEG2000 is the new international standard for still image compression. Quantization is the element of lossy compression systems responsible for reducing the precision of data in order to make them more compressible. JPEG2000 offers several different quantization options. The simplest form is scalar quantization using a fixed dead-zone. Nevertheless, this quantizer, based on a linear quantization, does not refer to human perception. So, the quality of compressed images can be remarkably improved if thedictates of the Human Visual System (HVS) requirements are followed. To achieve this goal, our strategy is to employ the wavelet structure to enable the visual masking properties to affect compression with an adaptive behavior. A combination of both intra channel and inter channel masking property is then performed. This scheme controls the step-size and the width of deadzone in the quantization process of a standard compatible compression chain. Then, we show with subjective assessment experiments that the quality of reconstructed color images is significantly improved compared to the base line JPEG2000 compression scheme.



Bibtex:

@inproceedings{Eude431,
    author    = { Thierry Eude and Christophe Charrier },
    title     = { A masking model for scalar non-fixed deadzone quantization in lossy JPEG2000 compression },
    booktitle = { 23rd Picture Coding Symposium (PCS 2003) },
    address   = { Saint-Malo, France },
    year      = { 2003 },
    month     = { April 23-25 }
}

Last modification: Feb 18 2003 2:48PM by eude

     
   
   

©2002-. Computer Vision and Systems Laboratory. All rights reserved