点击次数：

论文类型：期刊论文

发表时间：2006-03-01

发表刊物：7th Symposium on Measurement Technology and Intelligent Instruments

收录刊物：SCIE、EI、CPCI-S

卷号：17

期号：3,SI

页面范围：565-571

ISSN号：0957-0233

关键字：stereo matching; projective invariance; wavelet transform; 3D measurement; NURBS

摘要：This paper presents a NURBS (non-Uniform rational B-splines) Curves based stereo matching algorithm for which Curves replace the edges extracted with a wavelet-based method. Curves in the left images are projected onto the right view and the stereo correspondence problem becomes simply the estimation of the similarity between these projections and the original curves in the right image. The projective invariance characteristic of NURBS Curves is effective in reducing false matches due to distortion of the perspective projection and image noise. Experimental results demonstrate that the proposed method is robust and effective in removing abnormal reconstruction data. The obtained 3D reconstruction shape is closer to the real one.