Computational Depth Reconstruction For 3DTV

So far,the two-dimensional image/video technology has become more and more mature,but the need of humans is not only limited to the image itself.People also hope to experience the immersed feelings,just like the real life.So,the stereoscopic TV technique,namely 3DTV appears,and has become a hot field.Depth information is one of the necessary conditions to reproduce the real 3D scenes.Existing depth sensing methods and technologies have many shortcomings,to some extent,in terms of resolution,completeness,accuracy,and convenience,which heavily limits the development of related applications such as 3DTV and machine vision.In this paper,to obtain high quality and high accuracy depth information based on the existing depth acquisition mechanism,a priori model of the depth signal is constructed,and the reconstruction theory and method is put forward.The results and innovations of this paper are as follows:1.An adaptive color-guided autoregressive(AR)model for high quality depth recovery from low quality measurements captured by depth cameras is proposed.The depth recovery task is formulated into a minimization of AR prediction errors subject to measurement consistency.The AR predictor for each pixel is constructed according to both the local correlation in the initial depth map and the nonlocal similarity in the accompanied high quality color image.The proposed scheme is obviously superior to the current mainstream methods?2.A lightweight multi-view imaging approach with Kinect,a handheld integrated depth-color camera,under the depth image-based rendering(DIBR)framework is proposed.First,a progressive edge-guided trilateral filter is proposed to fill missing areas of the depth map.In view synthesis,a low-rank matrix restoration model to inpaint disocclusion regions is proposed,fully exploiting the nonlocal correlations in images,and an efficient algorithm is devised under the augmented lagrange multiplier(ALM)framework.Experimental results show that our method restores high quality depth maps even for large missing areas,and synthesizes natural multi-view images from restored depth maps.3.A motion-assisted matrix restoration model for background recovery and foreground extraction from video clips is proposed.The backgrounds across frames are modeled by a low-rank matrix,while the foreground objects are modeled by a sparse matrix.The dense motion field is estimated by optical flow,and mapped into a weighting matrix which indicates the likelihood that each pixel belongs to the background.By integrating the motion information into the model,we can get the separation results by convex optimization.In terms of depth video restoration,the judgment of moving objects in scenes can increase the accuracy of depth recovery,and enhance the coherence of the video in the time domain.So it is important to recognize the foreground from the background of the video.This work is a basis for the follow-up research of high-precision depth video restoration.