为了提高图像融合评价的准确性,提出了多尺度PSNR的图像融合客观评价方法。该方法首先对标准图像和融合图像按同种方式进行多尺度划分,得到每个标准图像块与相应融合图像块之间的PSNR;然后将它们组合成多尺度PSNR;最后应用灰色关联分析作为图像融合结果的优劣评判标准。实验结果表明,这种评价方法考虑了图像像素的整体和局部灰度分布特性,能正确对融合图像质量进行评价,其客观评价结果与主观评价结果具有一致性,且评价方法具有较好的有效性和简洁性。
This paper proposed an algorithm based on multi-scale PSNR to improve the accurateness of image fusion evaluation result.It obtained many sub-images when divided the fusion image and reference image in different scales,and organized multi-scale PSNR vectors with all PSNR values between each sub-reference image and sub-fusion image.It obtained the evaluation result by grey relation analysis.Experiments show that the proposed method is more consistent with subjective evaluation results,and the proposed method performs well on validity and simplicity.