基于方向可调滤波器的含噪图像融合算法

Fusion Algorithm of Noised Images Based on Steerable Filters

提出了一种基于方向可调滤波器的图像消噪及融合算法,即使用方向可调滤波器及其Hilbert变换组成的正交滤波器组,分析两幅已配准源图的主要方向及局部方向能量,根据噪声点所具有的奇异性、无方向性,从而检测出图像噪声点,并运用局部中值滤波方法消噪;对于源图中的特征,采用局部方向能量作为权重进行加权融合.该方法能提高信噪比约50%～80%,并能很好地保留和显著增强源图中的有用特征.

An image denoising and fusion algorithm based on steerable filters is presented. A quadrature filter pair is designed, which consists of a steerable filter and its Hilbert transform, to analyze the dominant orientation and local orientation energy in registered images. By using this quadrature pair, speckle noise can then be found out according to their singularity and nondirection properties. And then a local median filter is designed to remove these speckle noises. Simultaneously, significant image characters are fused via weighted averaging method based on local orientation energy. Results show that the signal noise ratio could increase by 50% to 80%, and those significant image characters are preserved well and enhanced obviously.