An Image Information Entropy-based Algorithm Of No-reference Image Quality Assessment

With the development of multimedia technology and computer network after the21st century, image data has been widely used as a way to describe the exchange and description of information. Image quality affect the effective of image information, thus the image quality evaluation technology is an important research aspect of current information processing research filed.The method of image quality assessment can be divided into subjective approach and objective approach. The subjective method is evaluated in terms of the mean opinion score (MOS) while the objective method use an objective way, which adopt a mathematical model to calculate or predict image quality scores. Since subjective way is costly, this paper focuses on the objective algorithms. This work mainly includes the following:First, the classification of research filed and method was surveyed. Second this work introduced the main characteristics of subjective and objective evaluation algorithms. Finally, focus on the information entropy-based method, this work is divided into three parts:(1)This work analyzed and implemented an efficient general-purpose no-reference image quality assessment model that utilizes local spatial and spectral entropy features on distorted images (SSEQ algorithm). We found that SSEQ matches well with human subjective opinions of image quality, and is statistically superior to the full-reference (FR) IQA algorithm SSIM and several top-performing NR IQA methods.(2) A no-reference image quality assessment algorithm was proposed, which utilizes local spatial-spectral-gradient entropy features on distorted images (SSGEQ algorithm). By using a2-stage framework of distortion classification followed by quality assessment, we utilized a support vector machine (SVM) to train an image distortion and quality prediction engine. The algorithm based on human visual characteristics, the attention to the image edge texture area is greater than the flat areas, and gradient can well reflect the image in the smallest details contrast and texture characteristics change. The simulation results show that the algorithm had an apparent improvement than the original one. Specifically, at first in the distortion classification process, Gaussian blur, JPEG compression and JPEG2000compression distortion of classification accuracy greatly improved; and the algorithm has a good monotonicity, accuracy and consistency; final we plotted the distributions of subjective scores and objective scores on a2-D graph and also plotted the fitted curve on the same figure, We find that SSGEQ matches well with human subjective opinions of image quality.(3) SSGEQ algorithm used in railway real-time image monitoring system.Finally, we concluded the work of this paper and proposed my own ideas about the future development direction.