【摘要】 为了研究传统统计方法在彩色夜视融合图像质量评价的适用性,在分析各种统计方法的基础上,选择了“熵”作为客观评价方法,采用C-均值聚类方法和自动确定最佳颜色类数方法计算彩色图像熵。实验对10种不同场景的微光与红外图像采用4种融合方法产生的40幅彩色图像进行了熵计算和“感知质量”与“细节”主观评价。统计分析结果显示,“感知质量”与“熵”的相关性系数只有0.41,说明熵统计方法在彩色夜视融合图像质量评价上存在困难。对采用统计方法评价彩色夜视融合图像质量存在问题作了讨论。更多还原

【Abstract】 The aim of this paper is to study the practicability of statistical methods for assessing color fusion images of night vision.Based on analyzing statistical methods,the entropy was selected to assess color fusion images for night vision.The entropy of a color fusion image was calculated by the optimal color numbers automatically obtained with clustering method.Experiments of subjective assessments and entropy calculations were performed for 40 color fusion images generated by 4 kinds of fusion algorithms fusing visible and infrared images of 10 scenes.Experimental result shows that the correlation coefficient between perceptual quality and entropy is only 0.41,the entropy method is difficult to assess color fusion images for night vision.The inherent shortages of the traditional statistical method were discussed in the end.更多还原