Prediction of Visual Quality Metrics in Lossy Image Compression - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2020

Prediction of Visual Quality Metrics in Lossy Image Compression

Résumé

Images of different origin are widely used nowadays in various applications including medical diagnostic systems, remote sensing, etc. Due to modern tendency to improve imaging system resolution and increase image size, it has often become necessary to compress images before their storage and transferring via communication lines. Lossy compression is mostly employed for this purpose and an important task for it is to find and provide an appropriate compromise between compression ratio and quality of compressed data, in the first order, image visual quality. This paper considers an approach to predicting visual quality characterized by the metrics MSEHVSM or, equivalently, PSNR-HVS-M for the coder AGU based on discrete cosine transform (DCT). It is demonstrated that it is possible to estimate MSEHVSM in a limited number of 8×8 pixel blocks and then to predict this metric for the entire image for the considered coder. The influence of image content and the number of analyzed blocks is studied. It is shown that 500 or 1000 blocks are usually enough to carry out prediction with appropriate accuracy. © 2020 IEEE.

Domaines

Chimie
Fichier non déposé

Dates et versions

hal-02892268 , version 1 (07-07-2020)

Identifiants

Citer

S. S. Krivenko, F Li, V. V. Lukin, Benoit Vozel, O. Krylova. Prediction of Visual Quality Metrics in Lossy Image Compression. 40th IEEE International Conference on Electronics and Nanotechnology, ELNANO 2020, Apr 2020, Kyiv, Ukraine. pp.478-483, ⟨10.1109/ELNANO50318.2020.9088819⟩. ⟨hal-02892268⟩
30 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More