Title
Psychovisual and Statistical Optimization of Quantization Tables for DCT Compression Engines
Abstract
Abstract: The paper presents a new and statistical robust algorithm able to improve the performance of the standard DCT compression algorithm for both perceived quality and compression size. The approach proposed combines together an information theoretical/statistical approach with HVS (Human Visual System) response functions. The methodology applied permits to obtain suitable quantization table for specific classes of images and specific viewing conditions. The paper presents a case study where the right parameters are learned after an extensive experimental phase, for three specific classes: Document, Landscape and Portrait. The results show both perceptive and measured (in term of PSNR) improvement. A further application shows how it is possible obtain significative improvement profiling the relative DCT error inside the pipeline of images acquired by typical digital sensors.
Year
DOI
Venue
2001
10.1109/ICIAP.2001.957076
ICIAP
Keywords
Field
DocType
specific class,relative dct error,specific viewing condition,standard dct compression algorithm,case study,significative improvement,statistical approach,human visual system,extensive experimental phase,compression size,statistical optimization,quantization tables,dct compression engines,psnr,robustness,visual system,psychology,compression algorithms,transform coding,image classification,data compression,visual perception,pipelines,digital sensors,dct,statistical analysis,quantization
Computer vision,Pattern recognition,Computer science,Human visual system model,Discrete cosine transform,Transform coding,Robustness (computer science),Trellis quantization,Artificial intelligence,Data compression,Quantization (signal processing),Contextual image classification
Conference
ISBN
Citations 
PageRank 
0-7695-1183-X
13
0.93
References 
Authors
12
4
Name
Order
Citations
PageRank
Sebastiano Battiato165978.73
Massimo Mancuso224817.22
Angelo Bosco3505.22
Mirko Guarnera4536.59