Objectives: While the stereo display technology brings the user the visual experience, it also makes the viewer feel uncomfortable in different degrees. The research shows that the main factors causing visual discomfort include viewing environment, three-dimensional display equipment, human physiological and psychological conditions and three-dimensional content, among which the lack of comfortable three-dimensional content is one of the important factors restricting the popularization of three-dimensional imaging technology. Therefore, this work combined with the significant characteristics of human vision, the influence of contrast factors on visual comfort is quantitatively studied, and then the quantitative range of contrast that affects the visual comfort is obtained.

Methods: Based on the visual attention mechanism, this work quantitatively studies the influence of contrast on the comfort of stereo images through a large number of subjective experiments. Firstly, the stereo salient map is obtained by combining the plane salient map with the stereo disparity map, and the optimized stereo salient map is combined with the source image to obtain the salient stereo image. The experiment of eye tracker verifies the correctness of the salient region extraction by the stereo saliency algorithm; Then the contrast linear change of left and right views is analyzed, and the subjective experiment is carried out for the distorted image; the least square method is used to fit the experimental data, and the comfortable contrast matching map and difference map of the significant stereo image are obtained, and the parts of the map are analyzed. The results of contrast verification and test verification show that the comfort range of contrast obtained by this method is more consistent with the subjective observation results of human vision, and provides a more scientific, reasonable and simple quantitative standard for the production of three-dimensional content.

Results: The upper left corner and lower right corner in Figure 7 (a) are not comfortable areas, mainly because the contrast difference between left and right viewpoints is too large at this time, and it is difficult for both eyes to blend comfortably. Watching such stereo image or video for a long time is easy to cause visual fatigue, thus inducing eye diseases. At this time, the contrast of right viewpoint should be reduced or increased appropriately, so as to facilitate the fusion of human eyes and binocular to get comfortable image; The lower left and upper right corners in figure 7(a) are the non-comfort areas. Although the contrast of left and right viewpoints is close and easy to binocular fusion, because the contrast of both viewpoints is too low or too high, the quality of the fusion stereo image is poor, resulting in visual discomfort. When making a binocular stereo image, we should pay attention to the use of the adjustment function of the main eye image, for example, we can use the low contrast of the main eye view to suppress the high contrast stimulation of the other view, so as to generate stereo content suitable for human eyes to watch. In addition, the spatial frequency of the salient region is concentrated in the high frequency region because the salient region contains more detail and texture information than the non-salient region.The optical transfer function shows that the response of the human eye imaging system to different spatial frequency components is different. The contrast sensitivity function further points out that the sensitivity of the contrast of the human eye in the high frequency segment is reduced. To sense the same contrast, the actual contrast of the image needs to be increased. Table 2 further shows that the comfortable contrast area obtained is very consistent with human visual characteristics. In order to verify whether the comfortable contrast matching image and the difference image obtained in this experiment are universal, two source stereo images are selected to verify the results. The experimental results verify that the comfortable contrast matching map and the difference map of the threedimensional image obtained have good universality. The conclusion can be used to judge whether the image is comfortable quickly and directly according to the contrast of the left and right views of the three-dimensional image, which greatly improves the present situation of time-consuming and laborious subjective experiment. The experimental results show that the contrast difference thresholds of lift and right view change with different left view contrasts, and the contrast difference should not be too large. The maximum and minimum difference values are 1.97 and -2.40, respectively.

Conclusion: Based on the mechanism of visual attention, a large number of subjective experiments are carried out to quantitatively study the influence of contrast on the comfort of stereo images, which better reflects the visual feeling of human eyes, and the comfortable contrast matching and difference maps of stereo images are obtained. On the one hand, we can judge whether the stereo image is comfortable according to the experimental results, which makes up for the timeconsuming and laborious shortcomings of the subjective experiment; on the other hand, through the comprehensive and detailed analysis of the result map, we can provide reference and basis for the production and processing of the stereo content. Compared with the experiment without considering the visual attention mechanism, the comfort range obtained in this work is more consistent with the characteristics of human vision, and can provide a more reasonable and feasible quantitative standard for the production of three-dimensional content. However, there are many factors that affect the visual comfort of stereo images, which need to be further studied in order to improve the theoretical system.