Visual discomfort prediction and reduction for stereoscopic 3D contents

Abstract

It is generally agreed that a possible obstacle to the widespread deployment of 3D services is related to issues of visual comfort. The industry has taken several steps to address the concerns, particularly in terms of 3D content production guidelines and post-processing tools. However, the solutions provided often require labor-intensive and time-consuming processes. As a result, there is an increasing demand for the development of automatic tools to predict and reduce visual discomfort of 3D contents. This thesis aims at improving visual comfort of stereoscopic 3D content by analyzing its scene content characteristics and manipulating them. To this end, this thesis analyzes 3D content characteristics that negatively affect visual comfort and provides a method to predict the degree of perceived visual discomfort. Based on our analysis and visual discomfort prediction method, we propose novel techniques to improve visual comfort of stereoscopic 3D content. The main contributions of this thesis are mainly summarized as follows: (1) development of disparity features to quantify different discomfort causes of stereoscopic 3D contents, (2) objective visual comfort assessment metric to predict visual discomfort of stereoscopic 3D contents, (3) combined linear and non-linear disparity remapping method to improve visual comfort under the guidance of the assessment metric, and (4) combined disparity adjustment and crosstalk cancellation for reducing the perceived crosstalk associated with scene content characteristics in stereoscopic 3D displays. To demonstrate the proposed methods, subjective assessments were conducted with diverse stereoscopic 3D contents. The results showed that the proposed visual discomfort prediction method provided higher prediction performance than existing methods. In addition, the proposed disparity remapping methods was capable of improving visual comfort while maintaining the overall viewing quality. The above-mentioned our research effor...