基于二元树复小波域的高精度运动估计研究与应用  

Research and Application on High Precision in Motion Estimation Based on Dual-tree Complex Wavelets

作　　者：张媛[1] 张小红[1] 

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《小型微型计算机系统》2014年第2期407-411,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(11062002)资助;江西省自然科学基金项目(2010GZS0083)资助;江西省教育厅科技项目(GJJ13435)资助

摘　　要：离散小波变换中的位移可变性严重影响了小波域运动估计的精度,为了减弱在动态信号估计中该可变性的影响,构造出一种新的二元树复小波结构,并依据小波低频子带和高频子带对移变性敏感性不同的特性,提出一种优化运动估计方法.将此方案应用于视频序列中图像的运动估计之中,实验结果表明新算法有更高的估计精度,提前判定不动块,减少了不必要的搜索时间,降低了运算复杂度.预测帧的峰值信噪比(PSNR)比基于二元树复小波的值有所提高,对在空间复杂度和运动复杂度都较高的视频序列的编码更有效.The shift-variant property of discrete wavelet transform affects wavelet-domain motion estimation accuracy seriously. In or- der to weaken the influence of the variability of estimation in dynamic signal, constructing a new dual-tree complex wavelet structure. According to the sensibility difference between low-frequency sub-band and high-frequency sub-band of the wavelet, this paper puts forward a optimized motion estimation method. Applying this scheme in motion estimation of images in video sequence, results show that, the new algorithm has higher estimation precision. Judging the stillness block in advance can reduce the unnecessary searching time, and can lower the computational complexity. The peak signal-to-noise ratio （ PSNR ） of inter frame is higher than the value of based on dual-tree complex wavelets method,and the new method is more effective in coding of video sequence with both high space complexity and motion complexity.

关 键 词：位移可变性 二元树复小波 双高通滤波 运动估计 运动补偿 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]