An objective visibility threshold measurement method for asymmetric stereoscopic images

Abstract

Due to the high measurement complexity in subjective evaluation methods, objective strategy is an effective approach to measure the visibility threshold (VT) between left and right views in asymmetric stereoscopic images. In this paper, a novel objective VT measurement method for asymmetric stereoscopic images is proposed. The proposed method first extracts stereoscopic features from left and right images with singular value decomposition, and the relationship between stereoscopic features and quality score is established with support vector regression (SVR). Finally, the SVR model is used to test specific stereoscopic images, and the objective VT is measured by statistical analyses. The experimental results show that the VT is related to the content of stereoscopic image and the quality of left view, and when the PSNR of left view is larger than 35dB, the VT value can exceed 2dB.