Comfort assessment of stereo images considering edge objects

Many people often feel visually uncomfortable when viewing stereo images. In the field of stereo display, it is an important research topic to assess comfort of stereo images accurately. However, the actual stereo image scene is more complex, and it is ineffective to establish the objective comfort assessment model of those images only by few factors. The current related research has shown that edge objects have some impact on visual comfort. Therefore, in order to assessment the visual comfort more accurately, a stereo image comfort assessment model considering edge objects is proposed. The influence of edge objects were analysed, and features of them were analyzed and computed, such as parallax, edge conflict, color and contrast features. In addition, region of interest features were computed by considering the visual characteristics of human eyes. Finally, an objective assessment model of stereo images was constructed by the support vector regression(SVR) algorithm based on the these features. The experimental results had shown that the performance assessment indexes PLCC, SROCC and RMSE are improved by 0.9%, 1.16%, and 10.6%, separately, on the IVY dataset.