夜视融合图像质量主观评价方法

为建立通用、客观的融合图像质量评价方法，在分析图像质量主观评价方法基础上，研究了融合图像质量主观评价对象、评价条件、评价指标和数据处理等关键环节，结合目标探测识别与场景理解两个典型的视觉任务，提出了目标可探测性和细节可分辨性两个主观评价指标，并研究了图像整体感知质量与这两个指标的相关性。对3种不同场景的微光与红外融合图像，采用9种融合方法获得的189幅融合图像的统计分析结果显示，细节可分辨性和图像整体感知质量相关性好，而目标可探测性虽然和图像整体感知质量相关性较差，但在基于具体视觉任务的融合图像质量评价过程中，仍可作为有效评价指标之一。

Subjective assessment method of night vision fusion image quality

In order to build the general and objective fusion image quality assessment method, the key technique of fusion image quality subjective assessment were presented such as assessment object, assessment condition, assessment index and data processing based on the analysis to image quality subjective assessment method. Two subjective assessment indexes were provided including target detectable ability and details resolvable ability according to two typical vision tasks of target detection and scene understanding. Furthermore, the correlation was investigated between the whole perceptual quality with target detectable ability and details resolvable ability respectively. Experiments were performed for 189 fusion images generated by 9 kinds of fusion algorithms fusing visible and infrared images under 3 kind different scenes. Statistical result showed that the high correlation between the whole perceptual quality with details resolvable ability compared with the low correlation to target detectable ability. However, the index of target detectable ability was still effective and could be used in the actually vision task.