Research On Image And Video Quality Estimation

As the development of the digital technology,especially more and more application of multimedia,image and video quality estimation become more and more important which is also one of the most difficult in the image processing. Based on gray image quality estimation,some reach has been done on color image quality estimation and video quality estimation.As the beginning of the paper,image quality estimation is introduced.It can be divided into two groups: objective image quality estimation and subjective image quality estimation.For the shortcomings of the subjective image quality estimation,for example,MOS (Mean Opinion of Score),much reach is payed attention to the objective image estimation.There are three steps in the development of objective image quality estimation,original image quality estimation,improved original image quality estimation and perception based image quality estimation.Some typical assessments in several types of image estimation are selected and compared in this paper.And then,it is found that every method has its disadvantage. So muliti-linear regression analysis is used to integrate these methods into one which could has the advantages of all the methods.During the integration,an idea of image quality measurement is gotten and proved.In this idea,three pieces of image information are used which is also agree with the process of subjective quality estimation.Accordingly,three models which stimulate these information are given and so are the functions of the HVS to these models.After then,the functions are integrated into a equation which is used to measure the image quality.Some comparison with proposed methods is given.As the result,we find that this model is more agreeable,accurate and consistent with the subjective image quality estimation,MOS.All the above methods are proper to the distorted images which are of the same size with the original images.But there are some other processing which would change the size of the images,such as image zooming.So a method is given which is used in such kind of image processing.An estimation based on Hausdoff distance is gotten and introduced here.Now color images are used more and more popular.So many color space models are used in the digital image processing.In this paper,about 10 kinds of color space models are introduced.And several simple image quahty estimation methods are compared in these color space models.Then we can find that different methods have different appearance in different color space models for different distortion.With the development of image compress technology,video is used more and more often.For the complexity of the subjective video quality estimation,more and more researchers are pay attention to the video quality estimation.In this paper,two kinds of subjective video quality estimation are introduced.And then some kinds of objective video quality estimation are also introduced here. After that an edge distortion based video quality estimation model is proposed which is proper to the quality estimation of video sequence.