基于WTMM的多重分形图像去噪算法

A New Multifractal Image Denoising Algorithm Based on WTMM

提出一种新的多重分形图象去噪算法,讨论了基于小波极大模的多重分形谱估计算法。在此基础上推导了图像取噪声的谱移位算子。该方法没有对噪声的类型提出任何假设条件,而是通过定义一个变换算子对每一点的Hausdorf指数进行处理,使处理后的图象的Hausdorf指数接近于2,从而取得最佳效果。实验结果表明,该方法在去除噪声的同时可很好地保留了原始图像的纹理信息。

In this paper we present a new approach for multifractal image denoising.The estimation of multifractal spectrum is completed with wavelet local maxima modulus.The advantage of this algorithm is that it can calculate the multifractal spectrum effectively and simply.An image denoise operator is gived by2-microlocal analysis.The method does not make assumption for the type of noise or global smoothness of orginal data.The image is characterized by its multifractal spectrum.This method processes Hausdorf exponent of the image by defining a new transformation operation,so that the transformed image has almost sure Hausdorf exponent a little above2.Experiment results explain that the method leads to a smooth image while preserving the relative strength of the singularities in the images.