Real-Time Depth Recovery Based on Projected Speckles

In order to solve the problem that many mismatching points exist when recovering speckle images by means of traditional matching algorithms,a region growing matching scheme characterized by selecting seed points at grid vertices and growing progressively is proposed.In order to address the problem of low speed and omission of certain regions,a fast method to select seed points at grid vertices is proposed.In terms of growing strategy,this paper stepwise loosens the criteria of seed points selection and region growing to progressively grow regions.To overcome the influence of uneven-brightness of speckle images,zero-mean normalized cross correlation method is adopted as the matching operator.In order to avoid the redundant calculation of zero-mean normalized cross correlation method,integral image and the more computationally efficient expression of zero-mean normalized cross correlation method are combined to speed the calculation and ensure the real-time performance.Finally,we interpolate the disparity image and transform the disparity to depth according to the triangulation theory.The experimental results show that,the proposed approach can recover the depth information of speckle images fast and robustly.