Digital image coding system using pattern recognition

All kinds of compression coding system for digital images can be described by an universal approach while the digital image coding is regarded as pattern classification. According to it, a new coding system called the vector quantization with constraint is designed and implemented. The basic princip...

Full description

Saved in:
Bibliographic Details
Main Authors: Yang Xuejun, Huang Zailu
Format: Conference Proceeding
Language:English
Subjects:
Adaptive filters
Data structures
Decoding
Digital images
Hilbert space
Image coding
Pattern classification
Pattern recognition
Vector quantization
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:All kinds of compression coding system for digital images can be described by an universal approach while the digital image coding is regarded as pattern classification. According to it, a new coding system called the vector quantization with constraint is designed and implemented. The basic principle is that the coding is accomplished by quantization of vectors from small blocks consisting of nxn image elements. It is first performed to classify the vectors into a high level pattern class meaning a subspace in Hilbert space which represents inherent data structure of images, and then it is classified into a low-level pattern class by quantization of some parameters in the high-level pattern calss. The classifier design is based on the functional theory about Hilbert space. In order to reduce the degradation due to blocking effect, the system also uses a fine adaptive filter for post-processing. The experiment shows that this coding system can obtain satisfactory quality of the decoded image at 0.875 bit/pel bitrate and the coding speed is high due to the algorithm simplicity.
DOI:10.1109/ICSMC.1988.754262