Font Size: a A A

Research And Implementation Of Auto Stereoscopic3D System’s Content Generation Based On Image Domain Warping

Posted on:2016-10-17Degree:MasterType:Thesis
Country:ChinaCandidate:H C XuFull Text:PDF
GTID:2298330467979378Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
With the rise of3D technology and the demand of audience, the mysterious veil of auto stereoscopic is uncovered. Behind the emerging technology, we can’t overlook the shortage of auto stereoscopic source. Limited by the photographing apparatus and transmission mode leads to the content shortage. For a long period of time, it’s not universal. The purpose of this paper is to study and put forward a reasonable proposal, will achieve the real-time auto stereoscopic content generation from traditional stereo video.Traditional proposal of the conversion system about multi-view video is based on the depth image, our paper’s method is on the basis of image domain warping. Using the sparse disparity of input stereo video, multi-view virtual video is generated real-time by the system. Compared with the traditional method such as the DIBR, the multi-view virtual video doesn’t contain the hole. At the same time, using image domain warping means the complexity of the algorithm is unacceptable. The main reason is the pre-process of obtaining the sparse disparity and the image optimization consumes much time. This paper has made the improvement in the corresponding algorithm, and puts forward the implementation framework of the hardware. Finally we reduce the hardware resource consumption and achieve the real-time conversion using the multilevel pipeline processing architecture to reduce the processing time. The main conclusions are as follows:(1) A method integrated a variety of feature extraction is proposed in this paper to realize the video pre-processing, including filter structure suitable for the extraction of stereo video feature and matching architecture. It greatly reduces the hardware cost and extraction time. Based on the real-time detection, sparse feature points of video matching hardware platform then real-time calculates the visual saliency weight.(2) In the process of real-time conversion system, the global optimization is the bottleneck of the system and the hardware resource consumption. In this paper, optimization problems, by subdividing the mathematical deduction and multiple rounds of iterative fixed-point decimal hardware simulation to solve the bottleneck problem of the overall scheme of time.(3) For the terminal of3D different display scheme of3D video reconstruction different rendering methods, this paper realizes the design of auto-stereoscopic display based on inclined column mirror grating. Analysis of the optical principle of the cylindrical lenses with tilt column mirror grating RGB sub pixel arrangement algorithm, designed the system framework and function modules of the hardware implementation, and finally realizes the synthetic display of multiple viewpoint image.Finally this paper regards Xilinx Vertex V FPGA as the main controller, design and implementation of the HDMI interface of the3D real-time naked eye content conversion and generation system based on. The overall scheme is simple and low cost, suitable for promotion in the industry, has great academic value and practical significance...
Keywords/Search Tags:auto-stereoscopic, FPGA, image domain warping, convex optimization, multi-view rearrangement
PDF Full Text Request
Related items