Research On Perception-based Visual Comfort And Quality Assessment For Stereoscopic Videos

Since the prevalence of stereoscopic video technology,people's pursuit of quality of threedimensional visual experience service has gradually improved.However,there are two issues in stereoscopic video system,such as viewing comfort and picture quality,which are most closely related to the three-dimensional visual experience.The establishment of stereoscopic video or image visual comfort and quality evaluation model in line with the subjective perception of the human eye has become a key technology in optimizing the stereo video technology to enhance the three-dimensional experience.This paper focuses on the two aspects of visual comfort and visual quality,and researches the visual comfort and quality assessment of stereoscopic videos in combination with human visual perception.The specific research contents are as follows:(1)According to the influence of perceptual characteristics of human visual system on the perceived quality of stereoscopic images,a no-reference stereoscopic image quality assessment method based on binocular visual perception is proposed.Firstly,the convergent cyclopean images are constructed to simulate binocular fusion and binocular rivalry of human vision system.At the same time,the uncertainty maps and disparity maps are established by considering the stereoscopic perception characteristic of human vision system.Then,the convergent cyclopean images are transformed into the curvelet domain and the features are extracted according to the multi-channel characteristic of human vision system.Meanwhile,statistical features are extracted from the uncertainty maps and disparity maps through natural scene statistical analysis.Finally,support vector regression is performed to predict the objective scores of stereoscopic images by establishing the relationship between the features and the subjective scores.The proposed method is tested on the LIVE Phase?and Phase? databases,and the Pearson linear correlation coefficient values of the method are above 0.94,and the Spearman rank correlation coefficient values are close to 0.94,significantly better than state-of-the-art two-dimensional and threedimensional image quality assessment methods.It is indicated that the proposed objective assessment model has good consistency with human subjective visual perception.(2)The existing visual comfort assessment methods of reducing the difference between the visual comfort features and the subjective assessment values by regression model have the limitation of relying on a large number of samples.Thus,a stereoscopic image visual comfort assessment method via discomfort-aware group sparse coding is proposed combined with the discomfort perception characteristic of human vision.Firstly,the disparity features are extracted combined with potential salient discomfort region and spatial frequency by considering the visual attention mechanism and the binocular fusion limit.Then,the features are grouped according to the subjective scores,and a discomfort-aware dictionary is constructed covering different levels of discomfort degree by clustering.Finally,the group sparse coefficients are obtained by performing group sparse coding on the dictionary,and the objective visual comfort scores are predicted by weighting group sparse coefficients and the discomfort values of each group.For the IVY LAB and EPFL stereoscopic image databases,the proposed method has better performance than the state-of-the-art stereoscopic image visual comfort assessment methods in terms of prediction accuracy and generalization ability,and closer to the human subjective perception.(3)For the visual discomfort phenomenon when human viewing stereoscopic images and the influence of disparity on visual discomfort,a stereoscopic image visual comfort enhancement method based on two-stage disparity remapping is proposed.Disparity scaling can improve the visual comfort of stereoscopic images,but also the three-dimensional sense of scenes will be over weakened.Thus,a two-stage disparity remapping strategy is designed to control the degree of disparity scaling degree for less-comfortable stereoscopic images.This strategy consists of the global linear remapping of the disparity range and the local nonlinear remapping of the disparity in the extracted potential less-comfortable regions.In addition,an objective visual comfort prediction model is constructed by using the global and local features to automatically guide the disparity remapping process.The global and local features are extracted by combining the human visual discomfort perception characteristic.The experimental results on the IVY LAB stereoscopic image database show that,the proposed method can more effectively enhance the visual comfort of lesscomfortable stereoscopic images while maintaining the three-dimensional sense of scenes compared with state-of-the-art stereoscopic image visual comfort enhancement methods,thereby enhancing the overall three-dimensional experience of stereoscopic images.