A Novel Full-Search Vector Quantization Algorithm Based on the Law of Cosines Author:Jarno Mielikainen, Student Member, IEEE Source:IEEE SIGNAL PROCESSING LETTERS, VOL. 9, NO. 6, JUNE 2002, pp.175-176. Adviser :Dr. Chang, Chin-Chen Reporter :Liang, Min-Chen Date :03/04/2003
Outline INTRODUCTION PROPOSED ALGORITHM EXPERIMENTAL RESULTS CONCLUSIONS
INTRODUCTION Vector Quantization (VQ) Generalized Lloyd Algorithm (GLA) Full Search
Generalized Lloyd Algorithm (GLA) =Linde-Buzo-Gray(LBG) Algorithm
Full Search
PROPOSED ALGORITHM d:exact distance, dmin:current minimum distance, d*:estimate distance x:a fixed vector, :angle between vectors u and v, 1:angle between vectors v and x, 2:angle between vectors u and x,
EXPERIMENTAL RESULTS PC:500MHz Pentium III, Compiler:GCC 2.95.2 with –O3 optimizations and loop unrolling, C Program:Khalid Sayood’s LBG program, Multispectral Image:AVIRIS 256*256*32
CONCLUSIONS A third constant vector to reduce the codeword-searching area. The proposed algorithm performed better than the GLA in experiments. Moreover, the algorithm is very simple to implement and is effective for large-dimension vectors.