DocumentCode :
280009
Title :
Quasi-periodicity classified vector quantization
Author :
Zhou, N. ; Daskalakis, C.N. ; Gold, M.P.
Author_Institution :
Dept. of Electr. Eng., Manchester Univ., UK
fYear :
1990
fDate :
33004
Firstpage :
42401
Lastpage :
42405
Abstract :
Proposes an alternative coding scheme which is based on a new criterion as well as a new image model. The main goal for image coding should be to preserve the essential structure of lines or shapes. From this point of view the code-book design is focused on representation of edges, not only because of their perceptual importance but also because edges are the most basic transition elements which can be used to construct various waveforms. Taking into account the statistics of edges in images, the authors propose an 8-class, 16-vector code-book (two vectors per class) which results from a training sequence using the LBG algorithm (the training pictures used are `Boat´ and `Flowers´)
Keywords :
analogue-digital conversion; encoding; picture processing; LBG algorithm; code-book design; edges; image coding; image model; statistics; training pictures; training sequence; vector quantization; waveforms;
fLanguage :
English
Publisher :
iet
Conference_Titel :
Low Bit Rate Image Coding, IEE Colloquium on
Conference_Location :
London
Type :
conf
Filename :
190126
Link To Document :
بازگشت