Stereoscopic image quality assessment combining statistical features and binocular theory

Stereoscopic image quality evaluation is extremely significance as a performance evaluator of modern 3D display technology. Due to the complexity of human visual system (HVS) and the incomprehensive study of stereoscopic perception of human eyes, stereoscopic image quality assessment (SIQA) is still a challenging task. In this paper, combining binocular characteristics, we propose an efficient no-reference stereoscopic image quality assessment according to binocular adding and subtracting channels. Distinguished from other SIQA methods, which pay attention to complex binocular visual properties, the visual information which is closely bound up with image distortion from adding and subtracting channels to describe ocular dominance (alternate name binocularity) is proposed. To estimate the contribution of each channel in SIQA, a dynamic weighting system is proposed for binocular fusion according to local energy. Furthermore, quality awareness features based on multi-scale and multi-orientation are extracted to describe visual degradation. Comparing with existing methods, experimental results on public 3D image databases demonstrate the proposed framework achieves high consistent with the subjective quality scores. (C) 2018 Elsevier B.V. All rights reserved.