基于相位一致性的结构相似度图像融合质量评价方法

在红外(IR)图像和可见光CCD图像融合效果评价中，结构轮廓对人眼感知的影响非常重要，而相位一致性（PC）模型生成的PC图像很好地在红外(IR)图像和可见光CCD图像融合效果评价中，结构轮廓对人眼感知的影响非常重要，而相位一致性（PC）模型生成的PC图像很好地保留了结构轮廓特征，且不受图像亮度和对比度变化的影响。提出了基于PC的结构相似度（SSIM）图像融合质量评价方法；利用SSIM模型分别产生融合图像相对于IR图像和可见光CCD图像的SSIM图像；根据PC模型产生IR和可见光CCD各自的PC图像；根据人眼对轮廓和非轮廓区域的重视程度的不同，利用各自的PC图像对SSIM图像进行加权； 最终利用信息熵加权得到融合图像的评价指标。实验中，在对实验图像进行主观评价和多种客观评 价指标计算的基础上，首次计算了非线性相关系数和斯皮尔曼等级相关系数(SROCC)，最后统计分析实验数据。实验结果表明该方法较以往评价方法具有更好的主客观一致性。

Image Fusion Quality Assessment Method Based on Phase Congruency and Structural Similarity

Structure contour is very important to the human visual system in the quality evaluation of infrared (IR) and visible CCD fusion images. The phase congruency (PC) image, produced by the PC model, can preserve the structure contour feature of image very well, even if the image brightness and contrast have changed. Therefore, an assessment method based on PC is proposed for these considerations. Firstly, two images, the structural similarity (SSIM) images of the fused image relative to the IR and visible CCD images, are produced by the SSIM model, respectively. Then, two PC images of the IR and visible CCD images are produced by the PC model. Afterwards, the SSIM images are weighted by their corresponding PC images, considering the different visual sensitive levels of two different regions, contour region and residual region. At last, the information-entropy coefficients of the IR and visible CCD images are used as the weights to predict image quality assessment. The subjective and objective indexes of experimental images are extracted, and the nonlinear correlation coefficients and Spearman’s rank-order coefficients (SROCCs) are calculated for the first time. Finally the experimental data are statistically analyzed. The results indicate that the proposed method can present more subjective evaluation results more than the previous evaluation methods.