A Stereoimage Synthesizing Method For Autostereoscopic Display With Optical-plate

The autostereoscopic display with optical-plate is based on the multi-viewautostereoscopic display technology with parallax barrier or lenticular-lens. Putting theoptical-plate in front of the display, the left and right eyes of a viewer can see theparallax images by using the separating function of light of optical-plate to get good3Dvisual.As a software component of autostereoscopic display with optical-plate, thestereoimage synthesizing method is very important. Therefore based on the opticalcharacteristic of optical-plate and the structural characteristic of LED or LCD panel, animage synthesizing method for autostereoscopic display with Optical-plate is proposedin this paper. There are three steps in the proposed method. First, a pixel mapping whichis used to synthesize the3D image from multi-view images is calculated according tothe parameters of autostereoscopic display. Secondly, multi-view images are resizedbased on the demand of the derived pixel mapping. At last, the3D image is synthesizedfrom the resized multi-view images.In order to meet the needs of people in any position to get3D visual, the more thenumber of viewpoints of autostereoscopic display with optical-plate is，the better.However, since the screen resolution is a fixed value, the horizontal resolution of eachviewpoint image reduced quickly with the increase of the number of viewpoints. Byusing the traditional method of image resizing for each viewpoint image, we would getbad3D visual because of the disproportion of vertical resolution and horizontalresolution. In this paper a seam carving for content-aware image resizing method isproposed to resize the horizontal resolution of each viewpoint image, so that thestructure and parallax information of each viewpoint image can be protected.Seam carving for content-aware image resizing uses an energy function definingthe importance of pixels，and the energy function is composed of gradient and parallaxinformation. Seam which is calculated by dynamic programming is a connected path oflow energy pixels crossing the image from top to bottom. By successively removing orinserting seams we can reduce or enlarge the size of an image. For image reduction,seam selection ensures that while preserving the image structure, we remove more ofthe low energy pixels and fewer of the high energy ones. For image enlarging, the orderof seam insertion ensures a balance between the original image content and theartificially inserted pixels. These operators produce, in effect, a content-aware resizingof images. The experimental results show that the proposed method, which gets better3Dvisual, can be applied to LED or LCD autostereoscopic display with lenticular orparallax barrier of any angle, and can synthesize the3D image of arbitrary resolutionand viewpoint quickly and efficiently.