一种高效的基于模拟退火的LBG算法

Efficient LBG Algorithm Based on Simulated Annealing

针对传统矢量量化码书设计 L BG算法对初始码书敏感和在迭代过程中容易陷入局部极小的缺陷 ,结合模拟退火算法 ,提出了一种基于模拟退火的 L BG改进算法 ,并给出了退火过程中的扰动因子刻画、扰动策略选取、稳定性判据确定和温度下降策略等细节 .模拟实验结果表明 ,本文所提出的改进算法能够有效地回避对初始码书的敏感 ,同时在搜索性能和图像压缩后还原质量上都得到很好的改善 .

An improved LBG algorithm based on Simulated Annealing has been proposed in this paper according to the LBG algorithm's defect in that is sensitive to the original codebook and is easy to fall into the partial optimization during the searching. At the same time, the details of the disturbance factor description, disturbance strategy choice, stability verification and the cooling table are all provided in this paper. The experiment results indicate that the improved algorithm can avoid the sensibility to the original codebook efficiently, and improve performance for searching ability and image quality of decoding excellently.