Quantitative Research, Particle-based Clustering Algorithm Context

The order entropy coding based on universal source coding theory is widely used in image, video, biomedical information and other coding systems. In order to solve 'model dilution' problem when achieving the order entropy coding, the context model must be quantified which goal is to design an optimal Context quantizer for the Context quantization and to further improve data compression ratio.Through the Context quantitative study, the Context quantizer design principles and the analysis of traditional K-means clustering and PSO clustering in this essay, the Context quantization system based on KPSO clustering algorithm is designed. This paper can be divided into two aspects:the construction of Context Models and the Context quantization based on clustering algorithm.In respect of building Context models, the image is made the 6 layer, two-dimensional discrete wavelet transform based on the wavelet transform theory in this paper. After transform, the coefficient matrix is quantified evenly with a cut-off area, and then the quantization factor is broken down into four small character symbol streams which include important position in stream, symbol stream, the highest position and the remaining bit stream flow. Finally,only the Context model for an important positon flow constructed is analyzed.In terms of the Context quantifying based on clustering algorithm, the design principles and quantitative criteria of the Context quantizer are discussed firstly and the appropriate distortion metrics is selected in this paper. Then the traditional K-means clustering algorithm and KPSO clustering algorithm are introduced in details. Finally, the specific realization process of Context quantizer is introduced.The image data in this paper is from TUSC-SIPI image database. The experiment proved that in that case of the quantitative series is given, the Context quantizer is more superior based on KPSO clustering algorithm than K-means clustering algorithm and the entropy coding efficiency has also improved.