Research On Image Quality Assessment Algorithms

With the fast development of modern science and technologies in visual signal processing, more and more image and video communication systems have been produced. Though these visual communication systems greatly facilitate our daily life, it remains a great challenge to access the service quality of the visual contents. To measure the quality of the pictures by human observers is quite time-consuming and costly, and thus many researches go into developing an automatic image quality assessment method. According to whether the original image is available, the image quality assessment algorithms can be roughly classified into 3 categories: full reference (FR) image quality assessment, reduced reference (RR) image quality assessment and no reference (NR) image quality assessment. This paper proposes a full reference and a reduced reference image quality assessment method. Which are based the similarity measurement of the multiscale edges. The major contribution of the paper is as follows:A brief introduction of the Human Visual System (HVS), from both the biological and psychological points of view. Various aspects of the HVS has been referred to, such as the anatomical structure, the contrast sensitivity function, the movement of the eyes and some masking effects of the HVS.The detailed introduction of the image quality assessment algorithms according to the abovementioned FR/RR/NR categorization.The introduction of multi resolution analysis of image, with emphasis on the dyadic wavelet transform and the multiscale edge extraction of image.The full reference and reduced reference image quality assessment algorithms based on similarity measurement of multiscale edges.