Stereoscopic image transforms to autostereoscopic multiplexed image

Recently, 3D display system are getting considerable attentions not only from theater but also from home. 3D multimedia content is revolutionizing visual entertainment which people can feel more real and fun than color and sound of 2D. Lenticular autostereoscopic display is one of the 3D-TV that consists of following advantages. It improve 3D viewing experience, support wider viewing angles and multiple viewers, and no need for special glasses. In this paper we describe a new technique for generating Autostereoscopic multiplexed content. First, the feature points detector Speeded-Up Robust Features (SURF) was used to match the corresponding pixels between stereo image, and we integrated the Epipolar Line property with detecting process to measure the accuracy of result and scale the matching range down; these corresponding pair, moreover, could be applied to estimate the depth of point which will be used to establish the depth map of image through interpolation of segmentation. Second, in order to obtain the optimized result of disoccluded regions caused by wrapping, we pre-process the depth map through Gaussian filter and mean filter. Finally, we proposed a new method based on DIBR to fill the disoccluded regions with more accurately. The experimental results of our system showed an impressive quality of rendering with Lenticular autostereoscopic display.