Research On 3d Image Quality Assessment Algorithm Based On Human Visual Perception

With the high-speed development of multimedia technologies in recent years, digital image has been widely accepted by human beings due to its specialties, such as reality, vividness, and abundance. Moreover, recent advances in multimedia technologies cause an increase in the field of three-dimensional(3D) image since it could bring stereo perception and immersive visual experience to human beings. Thus, it has become the hotspot and mainstream aspect in the research fields of multimedia technologies. It is well-known that the image delivery chain for the whole image transmission system includes five stages: content production, storage, transmission, decoding at the received side and then display. However, any of stages may cause the drop of the 3D image quality, and the error generated at a certain stage may result in the error propagation in the following stages of the delivery chain. The objective quality assessment method measures the physical aspects of the 3D image and considers the physical aspects and psychological issues. Its goal is to design the algorithms that can automatically evaluate the quality of the 3D image in the perceptually consistent manner. Thus, it is an efficient way to evaluate 3D image quality and reflect the performance of image processing system. Since human beings are the final receivers for 3D image, thus human visual system(HVS) has the crucial function in measuring the quality of 3D image. Based on the visual characteristics involved in HVS, in our paper, we proposed three novel image quality assessment algorithms.The structure of our paper is organized as follows. Firstly, we introduce the basic visual properties that are related with physiological and psychological issues for human beings perceiving the visual perception in 3D image. Secondly, subjective quality assessment method and objective quality assessment method for 3D image have been introduced in detail. And then, the proposed 3D image quality assessment algorithms have been illuminated, which are briefly described in the following.1) 3D image quality assessment algorithm based on the structural information. It is an algorithm that is based on the physiological and psychological properties of HVS. It is established to measure the block boundary distortion caused by block artifact, structure distortion and texture distortion in spatial domain in each image plane produced by image view in terms of the depth map. Experimental results show that the proposed algorithm could achieve better correlation withthe subjective assessment scores compared with the existing 3D image quality assessment(IQA) algorithms.2) 3D image quality assessment algorithm based on Discrete Cosine Transformation(DCT) technology. The algorithm is fulfilled by calculating the local pixel-based distortions in frequency domain after the image view coded by DCT, combining the simplified models of local human visibility properties embodied in frequency domain and the disparity information of 3D image. As confirmed in our experiments, the proposed algorithm not only has a good consistency with the subjective assessment scores, and can better reflect the human visual perception to 3D image than some other algorithms.3) 3D image quality assessment algorithm based on depth map information. The proposed algorithm is established based on the “image view plus depth map”, one of 3D image modes. And it calculates the distortions occurred in the depth map and counts the differences of visual acuity for human beings perceiving different visual areas on depth map, combining the structure of image view in spatial domain. From the statistical analysis, it is concluded that the proposed algorithm has a good consistency with human visual perception to 3D image, and could effectively predict the impact for depth map information to HVS.The basic scientific problems involved in the above-mentioned research achievements stem from our actual multimedia technologies development and application. Hence, they have the features of foresight and challenging, and have the important theory significance and study value. It can be confirmed that these researches could provide study value to the development of 3D image technologies to some extents.