No-reference Image Quality Assessment Method Associated With Statistical Properties

With the development of technology,a large number of digital images and video data bring convenience to peoples life and work,which also becomes an important research object with important economic value.However,the quality of image data is usually uneven,which brings great challenges to related application and research.It is of great significance to evaluate the quality of image and video data and predict low-quality data.Image and video data are often subject to various disturbances in the process of collection,storage,compression,and transmission,which lead to the data distortion and reducing the quality of collected data.How to analyze the changes of data quality accurately is the key task for image quality evaluation.Studies have shown that high-quality natural image data will show highly consistent statistical trends.But data distortion will destroy the structure of the images.Characterizing such changes can be used to evaluate the quality of image.When evaluating the quality of related distorted data,the quality evaluation methods are divided into methods with or without references by whether providing the original clean data as a reference.However,there is usually no clean data as a reference in real situations.Hence,most evaluation methods are no-reference methods.For the statistical characteristics of naturalness,structure and color information of image data,this paper will analyze and verify the changes of these statistical characteristics in distorted and clean data.Then we design a no-reference image quality evaluation method based on statistical characteristics to analyze the evaluation results.Efficient image quality evaluation methods can provide real-time feedback on image quality with high practical value.For example,for real-time judgment of image quality changing during live video and remote conferences,operators can adjust signals and strategies in time.Based on the distance model through the variance selection scheme,the proposed method in this paper improves the IL-NIEQ method by adjusting the 480-dimensional statistical characteristic index of the original method to 60-dimensional,which improves the calculation speed of the method and basically outperforms the previous methods.The performance is similar to the state-of-the-art method and better than some previous methods.