基于分数阶微分的图像增强模板

根据二维数字图像具有自相关性,为了充分利用邻近像素点的信息,推导出基于分数阶Riemann-Liouville定义的模板系数,构造了八个方向的分数阶图像增强模板;同时引进信息论中熵的概念对图像增强后的纹理保留效果进行定量分析。实验表明,提出的分数阶微分图像增强模板与传统方法相比具有更好的增强效果,并有效保留了图像的纹理细节信息,对纹理具有特殊需求的应用具有一定意义。

Image enhancement masks based on fractional differential

To take full advantage of pixels around the target pixel, this paper constructed a new fractional differential mask, according to the characteristics of high autocorrelation about 2-D digital image. Firstly, it deduced the fractional differential coefficients on the basis of Riemann-Liouville definition, and then filled eight differently directional masks. Meanwhile, it introduced entropy in information theory to describe quantitatively the retention of image texture when enhanced image. Simulation results indicate that, compared with traditional methods, this algorithm can make a better image enhancement and retain image texture highly, especially for the important application of the texture.