夜视抗晕光融合图像自适应分区质量评价

针对夜视晕光场景中，高亮度晕光信息导致现有红外与可见光融合图像评价方法失效的问题，该文提出一种自适应分区的融合图像质量评价方法。该方法根据可见光图像的晕光程度自动确定自适应系数，并通过迭代计算可见光灰度图像的晕光临界灰度值，将融合图像自动分为多个晕光区和非晕光区；在晕光区由设计的晕光消除度指标评价融合图像的晕光消除效果；在非晕光区从融合图像自身特性、对原始图像信息保留程度以及人眼视觉效果3方面评价融合图像纹理色彩等细节信息的增强效果；通过对4种不同抗晕光算法的融合图像进行评价分析，甄选出9种客观评价指标构成夜视抗晕光融合图像质量评价体系。不同夜视晕光场景下的实验结果表明，所提方法能够全面、合理地评价红外与可见光融合的抗晕光图像质量，解决了融合图像晕光消除越彻底客观评价结果反而越差的问题，也适于评判不同抗晕光融合算法的优劣。

Quality Evaluation of Night Vision Anti-halation Fusion Image Based on Adaptive Partition

To solve the failure of existing evaluation methods of infrared and visible fusion image caused by high brightness halation information in night vision halation scene, a novel fusion image quality evaluation method based on adaptive partition is proposed. In this method, the adaptive coefficient is automatically determined according to the halation degree of visible image, and then, the fusion image is divided into halo regions and non-halo region by iterative calculation of the critical halation gray value. In the halo region, the effectiveness of halation elimination is evaluated by halation elimination index designed, while in the non-halo region, the enhancement effect of detailed information such as texture and color is evaluated from three aspects including: characteristics of fusion image itself, retention degree of original image information and human visual effect. Based on evaluation and analysis of fusion images obtained by 4 different anti-halation algorithms, nine objective indexes are selected to construct a quality evaluation system of night vision anti-halation fused image. Experimental results in different night vision halation scenes show that the proposed method could evaluate anti-halation image quality of infrared and visible fusion comprehensively and reasonably, and could solve the problem that the more thorough halation elimination of fusion image, the worse objective evaluation results. This method could also be suitable for evaluating merits and demerits of different anti-halation fusion algorithms.