摘要
提出一种结合小波变换与图像分割的立体匹配算法。该算法利用双树复小波多通道提取立体像对带通相位信息作为匹配基元,求取初始视差场。针对单独通道提出一个相位匹配程度方程式,将视差计算归结为求解该方程的极大值。结合图像分割与视差平面拟合的方法对初始视差场进行修正,从而得到精度较高的密集视差图。实验结果表明,该算法结构简单,能快速有效地产生密集视差场。
This paper proposes a stereo matching algorithm combing wavelet transform and image segmentation. Multi-resolution phase acquired from Dual-Tree Complex Wavelet Transform(DTCWT) is used to get initial disparity map. For each scale, an equation is presented to measure the matching degree. Initial disparity map is amended into a high-resolution disparity map via color segmentation and disparity plane estimation. Experimental results show that the method is simple, and can achieve high accuracy and efficiency.
出处
《计算机工程》
CAS
CSCD
北大核心
2010年第17期210-212,共3页
Computer Engineering
基金
国家"863"计划基金资助项目"高可信网络立体电视应用系统关键技术研究"(2008AA01A319)
关键词
视差估计
相位匹配
双树复小波变换
图像分割
disparity estimation
phase matching
Dual-Tree Complex Wavelet Transform(DTCWT)
image segmentation