Research On An Improved Image Synthesizing Algorithm For Lenticular-based LCD Auto-Stereo Display

Compared with the flat panel display, stereo display technology can provide more comprehensive scene information, a realistic sense for the audience.Therefore, human being has always been to pursuit to reproduce the natural visual.Previous three-dimensional display technologies need a variety of glasses and other aids.But, stereoscopic display without wearing any gears allows us to enjoy free the visual impact of three-dimensional images in many locations,which greatly promotes the stereoscopic display technology.Because of its simple manufacturing process,high brightness,multi-view images, lenticular-based LCD auto-stereo display system has become today's research focus. After comparison of its advantages and disadvantages of multi-view image synthesizing algorithms,we do research on the image synthesizing algorithm of lenticular-based LCD auto-stereo display systems.The key work of this paper is as follows:Aiming at the Gibbs phenomenon and poorly real-time problem of digital filtering anti-aliasing algorithms in the lenticular-based LCD auto-stereo display system,this paper proposes an anti-aliasing algorithm,which is based on the principle of Full-Scene Anti-Aliasing(FSAA).According to Nyquist-Shannon sampling theory and the principle of FSAA,in order to increase the sampling rate,we firstly upsample the multi-view images under proportion, then downsample the images under non-proportion to achieve the anti-aliasing effects.Aiming at the formula's failure at the edge of lenticular, serial number's misplacement and redundant computing problems of the sub-pixel mapping algorithm, we improve the formula of the PHILIPS'sub-pixel mapping algorithm.In order to solve the failure problems at the edge of lenticular, we add an appropriate value to the formula, to enhance the formula's versatility.Meanwhile, the computer's internal floating-point error can cause the serial number misplacement, so we calculate a part of N viewpoints with the original formula, and the other viewpoints are calculated by the previous serial number.In order to validate the algorithm, in the experiment we use 3D scene extraction method for multi-view 3D raw material which can produce the highest 3D effect and quality. We use the 3D animation software---3DS Max.In order to compare with the digital filter algorithm,we use the Matlab platform to evaluate the algorithm from the aspects of visual effects,execution time,PSNR, SNR and others.Experimental results show that our anti-aliasing algorithm can eliminate the Gibbs phenomenon which is caused by digital filter algorithm,and also can effectively improve the algorithm in real time.The disadvantage is that our algorithm need more space.Meanwhile, theoretical analysis and experimental results show that the improved algorithm can solve the serial number misplacement's problem,while greatly reduce the redundancy computing in the mapping formula, improves algorithm in real time, and it can be easily implemented in hardware.