Single Camera Auto Stereoscopic Display Design

Autostereoscopic display is an important part in display research. Stereo video highlights the advantages of creating immersive experience for the audience, and providing a full range of media interaction. The traditional3D display system needs to carry auxiliary tools, so it is suffering to the eyes. Meanwhile, perspective is narrower, and equipment is expensive. Therefore, Research on a portable and real-time3D processing system with simple structure and faster processing capability is of great significance.The subject of this study is the project of IDC ASIA, Temasek Polytechnic in Singapore. Study and achieve a free3D display with five views. Focus on the display selecting, image preprocessing, object segmentation, deep computing, and disparity map rebuilding. FPGA is the core control processor, to achieve real-time acquisition, storage,3d processing and display system video. The major research work includes the following three aspects.(1) Analyze project-specific design requirements and select the hardware platform, then establish the basic hardware and software systems. Download into the FPGA for real-time online debugging. Analyze system resource usage. Divide image processing modules according to the FPGA design strategy.(2) Give and simulate five-view autostereoscopic display algorithms in MATLAB, then evaluate the algorithm effect, including noise reduction, image segmentation and image3D algorithm. Noise reduction processing is to simulate and compare the commonly used median filter and mean filter algorithms, as well as an adaptive weighted average algorithm, Image segmentation is to compared the Mahalanobis distance separation and parameter partition method, then improved an interactive, semi-automatic segmentation method. The3D algorithm is including depth algorithm, image reconstruction and sub-pixel arrangmen. Depth algorithm is based on the inverse operation of the reflectance map equation; the solution of the algorithm process is simplified through looking for the surface with minization brightness. Image reconstruction is achieved according to the depth of image projection principle. It is to restore the disparity map under the premise of the image depth is known.Sub-pixel arrangement is to design the arrangement program to match the design. Evaluate3D image with the method of cross-entropy and mean square difference.(3) Design the autostereoscopic display algorithm running in FPGA based on System Generator. Compute hardware resources consumed and embed design into the system, then download to FPGA to realize real-time acquisition, processing and display. The algorithm of hardware modules includes image pre-processing, image segmentation, as well as3D algorithm. Image pre-processing design is including the modules of mean filter, brightness/contrast controler, the Bayer format conversion, color balance, and Gamma compensation. Image segmentation and3D algorithm are simple processing, due to the high complexity algorithms. Segment the part which color is within the range of red, green and blue threshold and move objects horizontally, then use bilinear interpolation to make up the black hole;3D arrangement of the image uses the ROM block storing the disparity map, then achieve the output arrangement.