Image Quality Evaluation Method Based On Improved Structural Similarity

In recent years,image quality evaluation has become an important research direction in the field of image processing technology,and it has been rapidly developed.Image quality assessment methods are divided into two types according to the influence of human subjective factors,subjective evaluation and objective evaluation.Subjective assessment methods have many drawbacks in terms of evaluation effectiveness,including inefficiency,randomness,non-reproducibility,etc.The most critical is that they cannot be integrated in image processing.The objective evaluation is fast,the application field is wide,and the evaluation result is reproducible,which is not influenced by subjective factors.According to the original reference image information obtained by the objective evaluation method,it is divided into three categories: full reference,semi-reference,and no reference.In the process of image evaluation,the three statistical parameters of the information entropy,mean value and standard deviation of the image are all processed for the pixel values of the image,and have been well applied in image quality evaluation.For example,in the image quality evaluation method based on structural similarity,the image mean is used as the luminance information and the standard deviation is used as the contrast information;the spatial information entropy and the frequency spectrum information entropy are used as the referenceless image quality evaluation method based on information entropy.The eigenvalues are all better evaluated results,but the fuzzydistorted images are less accurate.At present,the study of image quality evaluation has become a very significant direction in the field of image processing.How to accurately and effectively evaluate the image quality is still a problem.The study of this paper is mainly based on the predecessors,on the basis of the quality of the wave image evaluation.Every year,the direct economic loss of China's marine disasters is about 5 billion yuan,and this number has been increasing continuously.Therefore,the marine construction project is inescapable and marine monitoring is an indispensable part.In the video surveillance system of Yangshan Port,digital images may have distortions in the process of acquisition,storage,transmission,display,etc.,resulting in degradation of image quality.In these processes,distorted images cannot meet the standards of offshore monitoring and waves.Detecting system requirements,it is therefore necessary to make an effective quality assessment of the images in the surveillance video.In order to improve the reliability and accuracy of the wave height monitoring results and reduce the errors in the classification of sea waves,it is of great significance to carry out a reasonable evaluation of the image quality of the collected sea wave images.This paper mainly improves the structural similarity part of the SSIM method and uses the masking effect for the image quality evaluation model.How to discriminate the flat area and the edge texture area in the image is the first problem to be solved;then the gradient is enhanced by the weighting function;finally the structure similarity theory with gradient weighting is obtained.The main research content is as follows:(1)The SSIM evaluation method is not sensitive enough to fuzzy-distortion image evaluation.To solve this problem,an improved image quality evaluation method based on structural similarity(Ws-SSIM)is proposed.The Ws-SSIM algorithm is more consistent than the SSIM algorithm in terms of consistency with the human eye.Based on the original algorithm,this algorithm not only has all the image information comparison functions in the original algorithm,but also has more gradient comparison functions than the original algorithm.Experiments on the ocean wave image database show that the Ws-SSIM algorithm is more in line with the human eye perception system and is superior to the traditional algorithm SSIM.(2)In order to further demonstrate that the improved algorithm is more suitable for the wave database than the traditional algorithm,the image matching algorithm based on the improved minimum distance method is also used to perform matching verification on the evaluated image.The experimental results show that the improvement is within the range of different evaluation values.The average matching errors of the images evaluated by the algorithm(Ws-SSIM)are lower than those evaluated by the SSIM algorithm.In this paper,the image database is constructed by using the ocean monitoring video of Yangshan Port,different types of noise are added to the images in the database,and the evaluation values of different noises are calculated using the SSIM,the improved SSIM,and the PSNR method respectively.The subjective sorting of different fuzzy images is performed,and the experiment shows that Ws-SSIM is more in line with human visual effects,SSIM is the second,and PSNR is the worst.The evaluation of images with different degrees of blurring indicates that Ws-SSIM is more suitable for evaluating fuzzy images.Experiments on LIVE data also show that the algorithm has good generalization performance.In order to further prove that this method is more suitable for oceanographic image of Yangshan Harbor than SSIM method,the method of image matching with improved minimum distance method is used to further verify the evaluation method.Experiments show that Ws-SSIM method is more suitable for oceanographic image of Yangshan Harbor than SSIM method.