Research On Crosstalk And Depth Reconstruction From A Single Image In The Autostereoscopic Display

Autostereoscopic display is one of the most popular development trends, without requiring any form of special glassed or other user-mounted devices. It provides viewers with convenience, especially for the children or someones wearing glasses. Crosstalk is the incomplete isolation of left and right image so that one image leaks into the other. It is a critical factor determining the image quality of stereoscopic displays. In this paper, a crosstalk measurement method and a crosstalk reduction method are proposed, based on the study of the crosstalk between the neighboring view images on the autostereoscopic display. Experimental results indicate that the proposed method is capable of improving 3-D image quality effectively on autostereoscopic display. Furthermore, in order to provide the depth imformation for stereoscopic display, a conditional random field(CRF) model with field of experts(FoE) as the prior for depth estimation from a single image is proposed in this paper.In order to relieve moiré and balance the resolution loss on horizontal direction and vertical direction, the slanted barrier is widely used nowadays. The experiment concerning the relationship between the slanted angles of barrier and crosstalk is difficult to realize, due to cost of barrier and lack of measurements. A model based on the subpixel judgment is proposed to measure the crosstalk, so that the quantitative research on the relationship between the slanted angles of barrier and the crosstalk is got. With the introduction of the crosstalk, the concept of effective resolution was defined to objectively evaluate autostereoscopic image resolution.On the autostereoscopic display, the crosstalk between the neighboring view images at each view image’s optimal viewing position is a critical factor affecting the 3-D display quality. How to eliminate the crosstalk as far as possible becomes important on the autostereoscopic display at present. The derived method based on the modified Schnorr-Euchner stategy is implemented to seek the optimal solution to this boxconstrained integer least squares(BILS) problem, such that the Euclidean distance between the solution and its target decreses subtantially. The result of experiment confirms that the proposed method can reduce the crosstalk between the neighboring view images at each view image’s optimal viewing position on stereoscopic display more effectively than existing methods in terms of mean square error(MSE) and peak signal to noise ratio(PSNR).Reconstruction depth map from the images is a common problem in the computer vision. In the traditional method, the simple probabilistic model, which cannot capture well statistic model of the ground-truth depth map and leads to the ambiguity in depth map. The high-dimensional FoE model is used as the prior in this paper, which obtains the sharp edge and the clear outline of the object in depth map corresponding to the image. In the traditional CRF model, it is assumed that the variance of estimation error is stationary. However, it is not usually true in the depth maps. In this paper, it is assumed that the variance of depth estimation error varies spatially in depth estimation model. The proposed method not only can enjoy the benefits offered by the flexible prior, but also has the advantage of making use of the probabilistic modeling to perform a posterior mean estimation, thus is less sensitive to the local minima problem as the MAP solution. Experimental results indicate that the proposed method outperforms state-of-the-art approaches in terms of RMSE-error and log10-error.