Current Issue Cover
面向水下图像集的一致性增强评价方法

孙晓帆,刘浩,张鑫生,吴乐明,况奇刚(东华大学信息学院)

摘 要
目的 由于水下图像通常存在模糊、失真、可见度低等图像质量退化问题,不断有新的水下图像增强算法投入到实际应用中。而目前的水下图像质量评价准则主要针对单幅图像进行评价,在对整个图像集的质量进行评价时,现有方法是采用某一质量评价准则的质量分数平均值作为指标,以平均值的高低来说明质量增强算法的优劣。但是,非一致性增强的质量分数平均值会随着图像集的变化而产生较大的波动,如果某一增强算法对小规模图像集都无法一致性地增强图像质量,那么当该算法投入到更大规模的图像集中时,以图像质量分数平均值作为评价标准,其局限性高、偏差大。为了解决上述问题,本文提出了一个更加具有普适性的水下图像质量评价
关键词
Consistent enhancement assessment for an underwater image set

Sun Xiaofan,Liu Hao,Zhang Xinsheng,Wu Leming,Kuang Qigang(College of Information Science and Technology,Donghua University)

Abstract
Objective Because the underwater images usually have the problem of quality degradation such as blurring, distortion and low visibility, an increasing number of underwater image enhancement methods have been put into practical applications. At present, each quality evaluation criterion mainly focuses on the single image. When evaluating the enhancement quality of a whole image set, the existing methods adopt the average quality score of a quality evaluation criterion as an indicator, and the enhancement algorithm is evaluated by the average score. However, the non-consistent average quality score will change with the image set and produce large fluctuations. If an enhancement algorithm cannot consistently improve the image quality score in a small-scale image set, the average quality score has some limitations and large error when the enhancement algorithm is put into a large-scale image set. To solve these problems, a more universal underwater image quality assessmentmethod consistent enhancement quality assessment (CEQA) for an underwater image set, is proposed in this paper.Method The proposed method can judge the consistency of the enhancement algorithm by comparing the difference of the quality score of the image enhancement before and after the image enhancement, and then by changing the weight proportion of the selected quality score difference, and unifying the fractional system, calculate the CEQA fraction of the enhanced image set. The concrete steps of this proposed method are as follows: Step 1: firstly determine an image set { I1、I2、I3、…、In} (n is the total number of images of the underwater image set), and then a quality evaluation criterion M is selected to evaluate the image quality of the original underwater image I1, and a quality score α1 of the original image I1 is obtained. Step 2: through the image quality enhancement algorithm A, the proposed method can process the original underwater image I1, and obtain the enhanced image I’1. Step 3: the proposed method uses the quality evaluation criterion M which is used in Step 1, to evaluate the quality of the enhanced image I’1 and get the quality score β1. Step 4: the quality score β1 minus the quality score α1 to get the fractional difference Q1. Step 5: successively carry out Steps 1 to 4 for the original underwater images I2、I3、…、In, and get fractional difference Q2、Q3、…、Qn, respectively. If the Q1、Q2、…、Qn values are all positive, it shows that the underwater image quality enhancement algorithm A, under the quality evaluation criterion M, can consistently enhance the quality of this underwater image set, and then proceed to Step 6; Otherwise, it is concluded that the quality enhancement algorithm A is not a consistent quality enhancement algorithm under these conditions. Step 6: select the maximum value of Q1、Q2、…、Qn as Qmax, minimum value as Qmin, and then find the average value of Q1、Q2、…、Qn as Qave. Step 7: by normalizing the average value Qave and the minimum value Qmin, and then adjusting its proportion, the effective value Ceff of the underwater image quality enhancement algorithm A for this image set is obtained under the quality evaluation criterion M. For the same underwater image set, under the selected quality evaluation criterion M, to evaluate different underwater image quality enhancement algorithms A1,A2,…,Am (m is the total number of the quality enhancement algorithms), the non-consistent quality enhancement algorithm cannot consistently enhance the image set under the quality evaluation criterion M; On the contrary, the consistent quality enhancement algorithm can consistently enhance the quality of the image set. When comparing several consistent quality enhancement algorithms: if the average value Qave is different, the quality enhancement algorithm, which has a higher effective value Ceff, has a greater enhancement strength and better enhancement ability; if the average value Qave is the same, the quality enhancement algorithm which has a higher effective value Ceff, has a better stability.Result In the "quantitative analysis of the mean value method", experimental results show that, after the image set is enhanced by the three image quality enhancement algorithms which are randomly selected, the average value of UCIQE and entropy is larger than that of the original image. However, there is a large number of single images whose quality score is lower than that of the original image. In the "extended application of CEQA method", it is concluded that by using the UCIQE evaluation criteria of selected underwater image set, the enhancement effect of the CLAHE-HSV algorithm is the best, and the inverse filtering algorithm is better than the filtering-guided dark channel defogging algorithm. A large number of experimental data shows that the above problems are well solved by our method, and the method provides an evaluation criterion for the quality enhancement algorithm of the image set. In the comparison experiment between the CEQA method and the mean value method, the experimental results show that, although the non-consistent quality enhancement algorithm has the highest mean value when the image set is small, its mean value is lower than that of the original image when the image set is enlarged. In other words, the non-consistent quality enhancement algorithm has an extensive or serious reduction of image quality. Before and after the expansion of the image set, the consistent quality enhancement algorithm can improve the image quality steadily, and thus the average value of the quality score is always higher than the mean value of the original image.Conclusion Through the "extended application of CEQA method" experiment, it is shown that the proposed method is feasible and can obtain effective experimental data by this methodology, so as to compare the advantages and disadvantages of underwater image quality enhancement algorithms. Through the “comparison between CEQA method and average value method” experiment, it is concluded that the proposed method in this paper is more accurate than the average value method, and effectively controls the large sample deviation. Therefore, this paper proposes a consistent enhancement assessment method for the underwater image quality, which provides a better evaluation criterion for underwater image quality enhancement algorithm in large-scale practical applications. The consistent enhancement evaluation method in this paper is better than the existing mean value method for the evaluation of an image set, and gives a quantifiable performance index for the new image quality enhancement algorithm. The proposed method in this paper has a guiding effect on the advantages and disadvantages of new image quality enhancement algorithm in the future. In addition, the formula of this method is simple, universal, highly flexible and easy to understand, it can be applied to various fields of image quality evaluation. The shortcomings of the proposed method are: this method has a higher requirement for robustness and stability of an enhancement algorithm. It is more suitable for those applications in the zero-fault-tolerance fields, and a reliable quality enhancement algorithm is selected for the application field with stringent performance requirements. For the common application requirements, the performance standard of this method is relatively high, and some algorithms without fluctuation cannot meet the requirements of this consistent enhancement assessment method. Underwater image enhancement technology still has much gap for development, and the enhancement performance of underwater image set needs more authoritative evaluation criterion. The further research direction of this work will focus on: making the method more fault-tolerant, and when facing a certain application, a good quality enhancement algorithm can be selected for different application requirements. Not only can an application select the quality enhancement algorithm under strict conditions, but also the application can change the screening conditions according to the specified requirements and get the specific experimental data and results.
Keywords
QQ在线


订阅号|日报