The Research Of Multi-view3D Video Processing Technology

At present the naked eye3D display has only one appropriate viewpoint, and theslight movement of the viewer position will cause crosstalk and affect the viewingexperience. To solve this problem, this paper presents a FPGA-based multi-viewprojection naked eye3D video processing method. The video sources will be dividedinto nine equal parts, and each of them will be enlarged as the original resolution. Thezoomed ones will be sent to nine projectors, then eight appropriate naked eye3Ddisplay views will be achieved after the following optical processing. Taking intoaccount the actual amplification effects, hardware limitations, the complexity ofalgorithm and other factors, the image magnification will adopt the bilinearinterpolation algorithm.The realization of this project includs new sources acquisition module, videoprocessing module, follow-up optical processing module and video processing moduleis the focus of this thesis. This thesis is based on the nearest neighbor interpolationalgorithm and bilinear interpolation algorithm theory. The image magnification ofMATLAB and FPGA have been introduced, the two algorithms are analyzed andcompared. This thesis is a combination of using MATLAB simulation and FPGAimplementation to process video, the following results and conclusions can be obtainedby studying:1. The two algorithms can be simulated by MATLAB and programming is easy toachieve, the calling of function and mathematical operations are more convenient, abetter simulation effect can be adopted, but it has a slower processing speed, and relieson the operating system.2.It takes for a long time for FPGA concrete realization during the whole process, thedesign of the circuit, welding, commissioning, programming are needed, the treatmenteffect can be restricted by many conditions, hardware design and programming canhave a significant impact on the treatment effect. The FPGA advantage is independentof the operating system, the effect can be modified by programming, it applies toreal-time processing for video as a result of fast processing speed. 3. In most cases, the nearest neighbor interpolation algorithm can get good results, butwhen the image contains pixel gray scale changes in the structure, the image transitionis not natural, traces of human processing can be found, mosaiceffect can be producedin the image, the edges of the image will appear jagged. Compared to the nearestneighbor interpolation, the bilinear interpolation algorithm can effectively avoid themosaic effect, the enlarged image is smoother and the transition is more natural, thevisual effect is relatively good, the arithmetic operation amount is small and easy toimplement. The new pixels are generated from neighboring pixels by the weightedaverage pixel, it will result in the loss of high-frequency components, resulting in theloss of image brightness.