Research On Real-Time Generation Of Light Field Display Content

Light field display technology has many advantages,such as full-color,continuous view,simple structure,no auxiliary equipment,no visual fatigue and so on.It has broad application prospects in entertainment,medicine,security,national defense and transportation.With the development of light field display technology,the generation technology of light field content is also gradually developing.Compared with the real scene acquisition,the computer generated light field image has high efficiency and small error,and can directly render the three-dimensional(3D)image of the virtual scene without the restriction of the external environment and other factors.In recent years,researchers have developed a variety of computer light field image generation technologies,which can adapt to a variety of data formats and display devices to obtain high-quality 3D images.However,the existing light field content generation technology still has some problems,such as low computational efficiency and poor compatibility.It is difficult to render large-scale scenes with complex animation and materials in real time,and there is a lack of real-time interactive 3D content.Aiming at the low computational efficiency of light field content generation technology,an in-depth study on computer-based light field content generation technology is made.According to the characteristics of light field content,several real-time light field content generation technologies and schemes are proposed.The main contents and innovations are as follows.(1)Real-time rendering method of depth-image-based multiple reference views for integral imaging displayTraditional DIBR can quickly generate multi-view images,but its inherent problem is that holes appear in the synthesized view.In order to quickly generate the content of light field while guaranteeing the image quality,based on rasterized rendering and viewpoint remapping technology,a method for generating the content of light field from multiple reference viewpoint depth images is presented.By increasing the number of reference viewpoints,this method improves the process of virtual viewpoint remapping,effectively reduces the generation of image voids during the generation of virtual viewpoints,and corrects the illumination error caused by the interpolation of viewpoints using the delayed coloring technology,which improves the image rendering quality.Qualitative and quantitative experiments are carried out to evaluate the feasibility of the proposed method.Experimental results show that the MDIBR method can achieve real-time integral imaging display with 80×80 viewpoints based on large-scale 3D data with a million points.(2)Real-time computer-generated integral image based on GPU-driven cross perspective rendering pipelineIn the traditional light field image rendering process,there are a lot of redundant calculations,which waste computing resources and seriously slow down the generation of light field content.In order to reduce computational redundancy and make full use of limited resources,A GPU-driven CGII method based on a cross perspective(GDRP)is presented,which optimizes the process of light field rendering in three ways.First,using cross-perspective projection technology,the rendering order is converted from viewpoint-by-viewpoint rendering to lens-by-lens rendering,which reduces the number of cameras.Second,with the geometry instancing technique,the rendering task is completed in only one pass,and the times of instancing are decreased,which reduces unnecessary data transfer between CPU and GPU.Third,the vertex culling in CPU and GPU are used to reduce the number of vertices fed into the rasterization stage.In the end,the GDRP is implemented and tested in several 3D scenes.All in all,the number of triangular faces processed by rasterization is reduced by 95%around,and the rendering speed of the image is improved.In the virtual scene,the frame rate rises above 60 FPS under the condition of 3840×2160 resolution,millions of vertices.(3)Parallel multi-view polygon rasterization for 3D light field display3D light field displays require samples of image data captured from a large number of regularly spaced camera images to produce a 3D image.Different from the single viewpoint generation,the multi-view generation has high information density,and it is required to explore more specialized generation methods specially designed for that purpose.However,the current 3D image generation algorithm with traditional single viewpoint computer graphics techniques is not sufficiently well suited to the task of generating images for the light field displays.A highly parallel multi-view polygon rasterization(PMR)algorithm for 3D multi-view image generation is presented.Based on the coherence of the triangular rasterization calculation among different viewpoints,the related rasterization algorithms including primitive setup,plane function,and barycentric coordinate interpolation in the screen space are derived.A soft rendering pipeline with GPU is designed and implemented.The PMR reduces the cost of producing the multi-view images by performing as many calculations as possible just once per sequence rather than once per viewpoint.In the algorithm,multiplications are transformed into accumulation operations as much as possible.Compared with traditional methods,it has higher parallelism and lower time algorithm complexity.To verify the proposed algorithm,a hierarchical soft rendering pipeline with GPU is designed and implemented.Several groups of images of 3D objects are used to verify the performance of the PMR method,and the correct 3D light field image can be achieved in real time.The frame rate rises above 60 FPS under the condition of 7680×4320 resolution,millions of vertices.