基于场景理解的双波段彩色融合图像质量评价

图像质量评价是双波段彩色融合处理算法及系统评价的基础，文中研究了一种可见光与红外彩色融合图像质量评价方法。提出了基于场景理解的图像感知质量PQSU 综合评价指标，选择三类典型场景彩色融合图像进行了主观视觉评价实验； 通过对已有评价指标与PQSU 综合指标的主观评价实验结果进行多元线性回归分析，建立了PQSU 的预测模型。结果表明:融合图像的场景颜色协调性与自然感高度相关；利用图像清晰度和颜色协调性可以有效地预测PQSU；针对不同场景类型，已有评价指标在PQSU 的预测模型中所占的权重有所不同，但预测模型的基本形式保持不变。文中提出的PQSU 及其预测模型为进一步发展融合图像质量客观评价模型奠定了基础。

Quality evaluation for dual-band color fusion images based on scene understanding

Image quality assessments are the basis for evaluations of dual-band color fusion algorithms and systems. A method of quality evaluation for visible and infrared color fusion images was explored. A comprehensive evaluation metric, image perceptual quality based on scene understanding (PQSU) was proposed, and color fusion images of three typical scenes were selected to perform a psychophysical experiment. The prediction model of PQSU was derived by multiple linear regression analysis of the experimental data for conventional image quality metrics and the proposed evaluation metric. The results show that the positive correlation between color harmony and color naturalness is very high. The variation of PQSU can be predicted effectively by color harmony and sharpness. In the three image categories, the proportional coefficients in prediction models for PQSU are different; whereas, the basic forms of Image quality assessments are the basis for evaluations of dual-band color fusion algorithms and systems. A method of quality evaluation for visible and infrared color fusion images was explored. A comprehensive evaluation metric, image perceptual quality based on scene understanding (PQSU) was proposed, and color fusion images of three typical scenes were selected to perform a psychophysical experiment. The prediction model of PQSU was derived by multiple linear regression analysis of the experimental data for conventional image quality metrics and the proposed evaluation metric. The results show that the positive correlation between color harmony and color naturalness is very high. The variation of PQSU can be predicted effectively by color harmony and sharpness. In the three image categories, the proportional coefficients in prediction models for PQSU are different; whereas, the basic forms of prediction models are unchanged. The proposed comprehensive evaluation metric and its prediction model provide a foundation for further developing objective quality evaluation of color fusion images.