Stereoscopic 3D crosstalk prediction

Abstract

A new metric is proposed to predict the perceived crosstalk using crosstalk level and original images rather than using the original and ghosted images. Our metric fully utilizes the color information of original image. We first extracted disparity map, color difference map and color contrast map from original stereo images. Then these information together constructs new metrics, which have high correlations with the perceived subject crosstalk scores. The prediction performance of these metrics are evaluated using various types of stereoscopic crosstalk images. Our approach by Pearson correlation and Spearman correlation achieves better evaluation performance than previous methods, which indicates that color information is one considerable factor to achieve high accuracy of crosstalk visible prediction. The experimental results demonstrate that the proposed metrics perform better than the previous ones.