Title :
Image compression using probabilistic winner-take-all learning
Author :
Osman, Hossam ; Fahmy, Moustafa M.
Author_Institution :
Dept. of Electr. & Comput. Eng., Queen´´s Univ., Kingston, Ont., Canada
Abstract :
The authors elsewhere have developed a new competitive learning (CL) algorithm called “the probabilistic winner-take-all (PWTA)”. This paper compares its performance to that of two well-known learning algorithms, namely, the optimal LBG and the frequency-sensitive CL (FSCL). It is shown that when all three algorithms were used to train a neural vector quantizer in an image compression application, the PWTA had a performance that was similar to the optimal one and that was significantly better than that of the FSCL
Keywords :
image coding; neural nets; unsupervised learning; vector quantisation; PWTA; competitive learning; image compression; learning algorithms; neural vector quantizer; probabilistic winner-take-all learning; Algorithm design and analysis; Bandwidth; Councils; Decoding; Distortion measurement; Frequency; Image coding; Neural networks; Power capacitors; Vector quantization;
Conference_Titel :
Circuits and Systems, 1994., Proceedings of the 37th Midwest Symposium on
Conference_Location :
Lafayette, LA
Print_ISBN :
0-7803-2428-5
DOI :
10.1109/MWSCAS.1994.518973
