Study Of Objective Quality Evaluation Algorithms Of Stereo Image

Recently, the technology of stereo imaging has caused wide concern by domestic and foreign scholars. However, stereo images will be inevitably introduced some distortion in those generation, processing, storage, transmission and display stage. The distortion of stereo images quality, the deficiency of stereo sense, the defect of viewing comfort level has become constraint factors on generalize and popularization of stereo products. How to evaluate stereo image quality effectively has become a focus in the relevant field in china and abroad. So far, there is no uniform standard in the field of objective assessment technical. Therefore, it is of long-term significance to establish evaluation mechanism can reflect subjective feeling accurately.A lot of existing stereo image quality objective assessment methods have been researched. In this paper, two stereo image quality objective assessment methods based on a full research of the human visual system and the basic principle of stereoscopic imaging, have been proposed. These two methods assess the stereo image in two aspects, stereo image quality and stereo sense. In the first method, the human visual system model has been built based on human visual characters. The reference and distortion images through the visual system model have been computed, resulting in an improved signal-to-noise ratio index. In the second method, an improved structure similarity evaluation method have been proposed because structure similarity has a defect cannot evaluate the noise and fuzzy distortion stereo image quality accurately. Therefore, structure similarity function have been introduced into the frequency domain and the average structural similarity index have been computed based on weighted average using texture information coefficient and spatial location influence coefficient as weight. As for the stereo sense, two objective assessment methods of stereo image quality have been used to get stereo sense index, combining with the absolute difference map based on the analysis of the principle of stereoscopic imaging.In order to verify the validity and reliability of the two stereo image quality objective assessment methods, this paper computed five correlation metrics and computational complexity between 13 objective assessment methods include two proposed methods to compare and analyze aiming at the common compression, noise and fuzzy distortion in the practical application. The experimental result showed that two proposed methods is in accord with the sense of human vision, while it can reflect the stereo image quality and stereo sense well. The improved signal-to-noise ratio method provided a preferable tradeoff between validity and reliability, while the improved structural similarity method needs to reduce the computational complexity to improve practicability. In a word, these two methods has certain reference value for the development of stereo image quality objective assessment technology.