Light Field Compression Based On New Video Coding Technique

Traditionally the input to a 3D graphics system is a sceneconsisting of geometricprimitives composed of different materialsand a set of lights. Based on this inputspecification, the rendering system computes and outputs an image. An importantproblem is that we require modest computational resources and are thus suitable forrealtimeimplementation on workstations and personal computers. Image-basedrendering systems generate different views of an environment from a set ofpre-acquired imagery. Recently a new approach to rendering has emerged:image-based rendering technology. Light-field rendering (LFR) has been proposed asan alternative rendering technique. Unfortunately, LFR requires tens to hundreds ofthousands of images to ensure photorealistic rendering results from any viewpoint.Compression is necessary.There are many method for it. Currently, the most efficienttechniques for light field compression use disparity compensation,analogous tomotion compensation in video compression. In previous work with disparitycompensation, a single disparity value is specified for a block of 8X8 pixels. In thispaper, we describe a multi-hypothesis scheme and divide the image to 4X4 blockwhich is based on H.264/MPEG-4. Our experimental results show an improvement ofimage quality and increase of PSNR.