Research On The Improved Vector Quantization Algorithm In Image Encoding

This paper has a thorough research of vector quantization (VQ) which is one of the most important image compression technologies. First, this paper gives the detailed introductions of basic principle, academic concept, key technologies, and development status about VQ. Then it has a deep research and exploration on the two key technologies of VQ : Codebook Design and Codewords Search.In the part of introduction about VQ, first it introduces the basic principle and academic concept briefly. Then it continues to give a detailed description of the three key technologies: Codebook Design, Codewords search, Indexes Assignment and their development statuses. At last, it gives all the types of current VQ systems.In the part of Codebook Design, it gives the optimal conditions of VQ system, which is the guiding principle in all the process. Then based on those optimal conditions, it describes the LBG algorithm in detail, which is the classic algorithm of Codebook Design or even the whole VQ and includes the principle, steps, advantages and disadvantages. Next, it introduces several improved LBG algorithms: the separated means algorithm and the sorted separated means algorithm. At last it proposes a novel improved LBG algorithm: an improved codebook design algorithm based on the sorted separated means algorithm. This algorithm takes the improve measure that puts the farthest input vector from the current codebook into the unreasonable cells to reduce the number of unreasonable cells. And as a result, it makes the division of vector space more reasonable. The digital emulation proves that this algorithm can get a better codebook.In the part of Codewords Search, it describes several fast codewords search algorithms in detail as bellow: full search algorithm, partial distance search algorithm, absolute error inequality algorithm, equal average nearest neighbor search algorithm, equal-average equal-variance nearest neighbor search algorithm, and the fast codewords search algorithms based on subvectors. Then this paper proposes a new codewords search algorithm: an improved codeword search algorithm based on subvector features. This algorithm uses the means and variances of input vector and its subvectors more effective. By the equation transformation, it makes some inequation criterions which are more effective. So it narrows the searching field and improves the searching efficiency. The digital emulation proves that this algorithm is better than others.