| Summary: | 碩士 === 國立屏東科技大學 === 資訊管理系 === 94 === Multimedia data transmission and storage are very important topics in information technology applications nowadays. Data compression is the key element to achieve these goals. In this thesis, a modified self-organizing tree algorithm is proposed, which is a binary tree searching method. We embed not only the dynamic path selection method to reduce the tree searching bias, but also the concept of ACO algorithm to dynamically change the threshold value in the traversed nodes listed in the searching path, which is utilizing the searching appropriate centroid process performed via each training vector. Depicting the similarity between each inner node and its child nodes of the tree structure progressively. As a result, hierarchical clusters will be constructed and this clustering rule can be used in encoding step of vector quantization to achieve the image compression.
In addition, we compare the proposed method with TSVQ, GLA/LBG, S-TREE and GATSM algorithm in image quality and codebook training time. The simulation results show that the proposed method outperforms those tree-structed algorithms, such as TSVQ, S-TREE in reconstruction image quality and takes less training time than those in LBG algorithm and GATSM algorithm.
|
|---|
