Quality Assessment for High Dynamic Range Stereoscopic Omnidirectional Image System

This paper focuses on visual experience of high dynamic range (HDR) stereoscopic omnidirectional image (HSOI) system, which includes such as HSOI generation, encoding/decoding, tone mapping (TM) and terminal visualization. From the perspective of quantifying coding distortion and TM distortion in HSOI system, a "no-reference (NR) plus reduced-reference (RR)" HSOI quality assessment method is proposed by combining Retinex theory and two-layer distortion simulation of HSOI system. The NR module quantizes coding distortion for HDR images only with coding distortion. The RR module mainly measures the effect of TM operator based on the HDR image only with coding distortion and the mixed distorted image after TM. Experimental results show that the objective prediction of the proposed method is better compared some representative method and more consistent with users' visual perception.