Objective Video Quality Assessment Based On Structural Similarity

With the rapid development of mobile Internet and video communication has gradually become an important carrier of information transmission and communication in people's daily life.It is now the explosive growth of social media.For example,micro-blog and We Chat in the presence of a large number of video communications.The quality requirement of video image is improved in communication.However,in the process of acquisition,coding and transmission,video images may be distorted.At this time the need to present in front of the video quality testing,in order to optimize the system to provide a better visual experience.From the user perspective,subjective video quality evaluation is more real.But this method requires more manpower and material resources,low efficiency,and can-not be applied to the actual evaluation system.The objective video quality evaluation method can't fully meet the subjective visual experience in the evaluation process.It is an important task to study the objective video quality evaluation which is more in line with the user's real experience.The human visual system can extract the structure information in the visual scene.The method of structural similarity is used to evaluate the quality of video image,which is mainly used to measure the distortion of video image by changing the structure information in the scene.In this paper,we study the theory of structural similarity and the method of application in detail,and on the basis of this,we make two improvements.Firstly,fuzzy image evaluation of structural similarity algorithm serious distortion effect is not ideal.We use the histogram to represent the structural features of single frame image.Showing concentration histogram with the fuzzy degree of single frame image and combined with motion estimation to evaluate video quality.Secondly,according to the structural similarity algorithm in the evaluation process does not consider the region of interest.The video image is divided into regions of interest and region of no interest.Because the edge structure is more abundant in the region of interest.So the edge structure similarity method is used in this area.In the region of no interest using structural similarity method.Finally,twothe evaluation results are weighted.It can distinguish the important information in the video image and has a good evaluation effect.The experimental results show that the method based on the improved structure information and motion estimation has better performance than the single scale structure similarity method in the evaluation of severely distorted video.The evaluation method based on visual region of interest is more consistent with the characteristics of human eyes than other algorithms and the consistency of subjective and objective evaluation is better.