Stereoacuity-guided depth image based rendering

This paper presents a novel method to improve the stereopsis and visual comfort of 3D images based on an idea that apply the stereoacuity prior to the process of synthesizing a virtual view by DIBR (Depth image based rendering) technique. DIBR technique is widely used in 3D film and TV programs. However, an emphatic issue is that the stereo video has a weak stereo-effect and viewing discomfort since without considering the true depth perception experience of human visual system (HVS) when directly observes a natural scene. Stereoacuity is a representation of the depth perception nature of HVS and the novel method will explain how the stereoacuity prior can be used to guide and optimize the process of DIBR. In this paper, the presented method does neither refer to the intrinsic nature of HVS nor try to construct an accurate computational model of stereoacuity. Instead, it is accomplished by using the contrast stretch method which is a nonlinear image processing operation to approximately simulate the distribution rule of disparities around the focal plane. Experimental results show that the proposed method provides a more precise depth perception around the focal plane, as is consistent with the nature of HVS, and thus the viewers are able to enjoy a much more comfort 3D experience.