IEEE Std 3333.1.2-2017 pdf download – IEEE Standard for the Perceptual Quality Assessment of Three-Dimensional (3D) and Ultra-High-Definition (UHD) Contents.
4. An overview of the standard 4.1 Quality assessment for 3D contents The separated perception of left and right views of stereoscopic 3D stimuli induces various binocular- relevant perception processes in the human visual system, which results in significantly different determination of the visual quality of the 3D stimuli from that of 2D stimuli. Indeed, existing perceptual image-quality-assessment models designed for 2D images generally fail to accurately predict quality scores of 3D images. To analyze and predict the degree of perceptual quality of a 3D content, it is necessary to understand the contents in terms of spatial and temporal characteristics that are based on existing psychophysical and statistical models of binocular visual perception. In addition, a perceptual quality- prediction model of depth-image-based rendering (DIBR)–synthesized views is suggested. Toward this goal, specific artifacts related to DIBR systems need to be considered. Mostly, the subjective assessment is inherited from what has been traditionally done for 2D subjective assessment as defined by ITU-P.910, ITU-R BT.500-13, and ITU-R BT.2021. 3 However, it is doubtful whether their results are reliable enough to be used as the reference because the viewing environment is quite different from 2D due to intensive immersion of a user wearing the glasses in dark lighting. Hence, to perform the 3D image and video assessments, new databases shall be required to cover the characteristics of human perception, display mechanism, viewing environment, and so on.
5. Quality assessment for 3D contents 5.1 General The interest in stereoscopic image quality has been increasing lately, which is stimulated by both entertainment industry and academic research (Lee and Lee [B20]). 4 In practice, stereoscopic images can be degraded by common processes, such as compression or transmission. Similar to monocular cases, there are three types of image-quality assessment: full-reference, reduced-reference, and no-reference quality assessment. In this standard, all these methods are explored in succession. 5.2 Full-reference quality assessment of stereoscopic images using disparity- gradient-phase similarity 5.2.1 General The flowchart of the 3D image-quality assessment method is shown in Figure 1. First, reference and distorted cyclopean images are generated by using the reference and distorted stereoscopic images based on a binocular combination/fusion model, respectively. Then, disparity maps of the reference and distorted stereo pairs, gradient magnitude maps, and phase congruency maps of reference and distorted cyclopean images are correspondingly extracted. Similarity measurements of the three features are computed, and then combined into the final stereoscopic image-quality index.