SVD-Based 3D Image Quality Assessment by Using Depth Information

Currently, as demonstrated in numerous researches and studies of evaluating the three-dimension (3D) image quality, two-dimension (2D) image quality assessment methods cannot be directly applied to measure the quality of 3D image. With the increasing demands of end-users to the visual perception in 3D image, it is necessary and urgent to propose efficient 3D image quality assessment methods. In this paper, a novel 3D image quality assessment method is proposed. In the proposed method, the image pixel blocks are firstly separated into different planes according to their depth values on the basis of the perception of human visual system (HVS). The singular value decomposition (SVD) mechanism is applied into different planes respectively. Then, the final results are calculated in terms of the global error, which is the distance of the distorted image deviated from the original image. To evaluate the performance of the proposed method, the popular LIVE 3D image quality database is utilized in our experiments. As shown in our experimental results, the proposed method has a better performance compared with other methods.