基于EMD的无参考图像清晰度评价方法

针对现有的灰度变化统计函数不能提供图像全频段综合评价的缺陷，提出了一种新的基于经验模式分解的无参考清晰度评价方法。图像经经验模式分解后，会产生多层本征模式函数图像和剩余图像。不同本征模式函数图像包含了不同频率区间的边缘、纹理信息。模糊图像和清晰图像在相同本征模式函数分解层上，表现出不同的灰度变化特性。通过统计各层本征模式函数图像的极值个数，作等权重加和，并利用整幅图像像素数进行归一化处理，构建了经验模式分解清晰度参数。仿真模糊图像和遥感图像的试验结果验证了该参数评价清晰度的有效性。

No-reference Evaluation Method of Image Definition Based on EMD

Current statistical functions of gray changes can＇t supply the comprehensive assessment of the whole frequency spectrum of images. In order to overcome this shortcoming, a novel no-reference evaluation methods of the image definition based on the Empirical Mode Decomposition （EMD） is proposed. After EMD, the image can be expressed as multi-layer Intrinsic Mode Function （IMF） images and the residue image. The dif- ferent IMF images include the information of edges and details which corresponds to the different range of frequency spectrum. At the same decomposition layer, blur image and clear image show different characteristics of gray changes. Firstly, the extreme values of each IMF layer are counted, then the number of each layer＇s extreme values are added together with equal weight. Finally, the summation is normalized using the whole number of image pixels. The result is the definition evaluation parameter based on EMD. The tests of simulation images and remote sensing images show the parameter＇s validity for assessing definition.