The increasing demand for image dehazing-based applications has raised the value of efficient evaluation and benchmarking for image dehazing algorithms. Several perspectives, such as inhomogeneous foggy, homogenous foggy, and dark foggy scenes, have been considered in multi-criteria evaluation. The benchmarking for the selection of the best image dehazing intelligent algorithm based on multi-criteria perspectives is a challenging task owing to (a) multiple evaluation criteria, (b) criteria importance, (c) data variation, (d) criteria conflict, and (e) criteria tradeoff. A generally accepted framework for benchmarking image dehazing performance is unavailable in the existing literature. This study proposes a novel multi-perspective (i.e., an inhomogeneous foggy scene, a homogenous foggy scene, and a dark foggy scene) benchmarking framework for the selection of the best image dehazing intelligent algorithm based on multi-criteria analysis. Experiments were conducted in three stages. First was an evaluation experiment with five algorithms as part of matrix data. Second was a crossover between image dehazing intelligent algorithms and a set of target evaluation criteria to obtain matrix data. Third was the ranking of the image dehazing intelligent algorithms through integrated best–worst and VIseKriterijumska Optimizacija I Kompromisno Resenje methods. Individual and group decision-making contexts were applied to demonstrate the efficiency of the proposed framework. The mean was used to objectively validate the ranks given by group decision-making contexts. Checklist and benchmarking scenarios were provided to compare the proposed framework with an existing benchmark study. The proposed framework achieved a significant result in terms of selecting the best image dehazing algorithm.

中文翻译：

---

一种基于 BWM 和 Group VIKOR 技术的图像去雾智能算法选择的新型多视角基准测试框架

对基于图像去雾的应用日益增长的需求提高了对图像去雾算法进行有效评估和基准测试的价值。在多标准评价中，考虑了不均匀雾、均匀雾和暗雾场景等几种视角。由于（a）多个评估标准，（b）标准重要性，（c）数据变化，（d）标准冲突，以及(e) 标准权衡。现有文献中没有公认的图像去雾性能基准测试框架。本研究提出了一种新颖的多视角（即非均匀雾景、均质雾景、和一个暗雾场景）基准框架，用于基于多准则分析选择最佳图像去雾智能算法。实验分三个阶段进行。首先是使用五种算法作为矩阵数据的一部分进行评估实验。二是图像去雾智能算法与一套目标评价标准的交叉，以获得矩阵数据。第三是通过综合最佳-最差和 VIseKriterijumska Optimizacija I Kompromisno Resenje 方法对图像去雾智能算法的排名。个人和团体的决策背景被用来证明所提出框架的效率。该平均值用于客观地验证群体决策环境给出的等级。提供了清单和基准测试方案，以将提议的框架与现有的基准研究进行比较。所提出的框架在选择最佳图像去雾算法方面取得了显著成果。