Research On Comfort Influencing Factors Of Stereo Image And Its Quality Evaluation Method

Stereoscopic display technology has brought a new visual experience to audiences,which has greatly improved the promotion and popularization of 3D products.Comfort stereoscopic display is the ultimate goal of the development of stereoscopic display technology.The visual discomfort that occurs while viewing stereo contents is an important reason hindering the development of 3D technology.Therefore,how to obtain comfort stereo contents according to the influencing factors that restrict the comfortable display of stereoscopic contents,and how to evaluate the stereoscopic image quality is one of the key issues to be studied in the field of comfort stereo display technology.The lack of comfort 3D contents greatly limits the promotion of stereo products.In order to produce comfort stereoscopic content in large quantities,we first need to find out which factors will affect the comfort of 3D contents,and then it is necessary to quantitatively give the comfort range of each influencing factor.The previous researches qualitatively studied the influencing factors affecting the comfort of stereo content,such as brightness,chromaticity,contrast,and disparity.In order to promote the production of 3D content,some researches obtained a comfortable quantitative range of various factors.However,these researches are based on the whole image and does not consider the visual saliency characteristics of human visual system.Therefore,this paper combines the visual attention mechanism to study the influence of contrast factor and disparity factor on the comfort of stereo images.Firstly,the visual salient regions of stereo images are segmented based on the visual attention mechanism,and then the contrast linear transformation method and the parallax translation method are used to obtain subjective experimental materials,and then subjective experiments are conducted.Finally,the least squares method is used to quantitatively obtain a comfort contrast range and a comfort disparity range.The comparison experiments show that the obtained comfort ranges of contrast and disparity are accurate and effective,which provide a more reliable quantitative standard for the production of comfort stereo contents.Stereo image quality assessment can not only predict if the 3D content can satisfy the requirements of comfort stereo display,but also can be used to evaluate theperformance of other 3D technical methods.Subjective quality evaluation is time-consuming,which leads the objective assessment algorithm a popular research topic in comfort stereo display field.Therefore,this paper proposes a no-reference objective assessment method for stereo image quality.Firstly,based on human visual system,a binocular fusion model is proposed,which can fuse the left and right views of symmetric and asymmetric distorted stereo images.Then the HOG and LBP features are extracted to represent the gradient and texture characteristics of the cyclopean image.Finally,a stacking based ensemble learning method is adopted.Three SVR learners are used to map the proposed features of the stereo image to its MOS value.A large number of comparative experiments show that the proposed method achieves high consistency with the subjective assessment,further more,it has a better performance on asymmetric distorted images than symmetric distorted stereo images.