| Vector quantization is an effective method of compression. It widely used in satelliteimage and remote sensing image compression because of its simple codec and highcompression rate.It also could be used in compression, storage and transmission of digital TVand DVD video. The development of vector quantization technical is aboult20years, Itgradually formed the three key research direction:codebook design,codeword search andindex assignment.And the codebook design is the most critical of the three. There are avariety of local optimal,near global optimal and global optimal codebook design algorithmbecause we only know about the necessary condition of optimal codebook in theory.It hasprominent defects to restrict this techniqueusing in many areas Althoughvector quantizationtechnology has many advantages. At present, the bottleneck of vector quantization is thespeed of codebook generation is not fast enough and the quality is not good enough orboth.Aiming at the improving generation speed and quality of codebook, two algorithms areproposed respectively"niche monkeys group multiply simulation algorithm" and "selfevolution glowworm swarm optimization algorithm".Genetic algorithm is a bionics algorithm which could effectively global search. Thearticle propose a new method of genetic algorithm which based on the niche ideas andconstraining the cross space based on the traditional codebook design of genetic algorithm ofdeviding training vector.The training vector could be classification with some randomnessand minimum distance in high-dimensional space based on the better reference codebook.Itcould improve the quality of codebook continuously. It also give the improvement ofconvergence of the algorithm. The experiment also show that the algorithm increase in thegeneration rate of codebook without reducing the quality of codebook.The self-evolution of firefly algorithm is to improve the codebook quality problems.Inthe codebook design of vector quantization, if the current codebook is not a global optimalcodebook, we could continue optimize it.Vector quantization is a searching extreme problemwith multi-peak of high-dimension. It makes a lot of good algorithms which solve single peakextreme problem well, couldn’t be used in the vector quantization.And firefly algorithm is one of them.The article make a careful analysis of defect in optimization process of codebookusing firefly algorithm. And proposed an improved algorithm.Through the dynamicadjustment of concussion coefficient,it increase the distance of firefly,reduce the parameter ofAttractive and also expand the search area which suppressing the algorithm convergenceexcessive phenomenon. The experiments show that the algorithm can improve the codebookperformance0.2-0.45codebook and more optimized.In the last chapter,it introduces the usage of codebook in image processing,which taksthe codebook generation algorithm as the representative described in this article.And make thegeneration codebook embedded into texture classification and digital watermarking in thirdchapter.The texture classification using K-view algorithm as a classification framework andalso use generated codebook using niche algorithm as texture primitives in classification.Itachieves good results.In the digital watermark,the codebook embedded in watermarkinformation directly and accept all kinds of attacks,the experimental results show that thegeneration codebook which proposed in chapter third have good qualityThe contribution of this paper on the vector quantization codebook design is to solvetwo bottlenecks which restrict the development of vecror quantization technique.Thesimulation algorithm of niche group multiply presents an fast genetic codebook generationalgorithm which does not depend on the schema theorem and building block hypothesis.Thegeneration speed of codebook greatly accelerated.Self evolution firefly algorithm realize thecontinuing optimization of existing codebook through controlling the parameter ofgravity,and also greatly inproves the quality of the codebook. |
PDF Full Text Request