红外图像的自适应混合双边滤波算法  

Adaptive hybrid bilateral filtering algorithm for infrared image

作　　者：余博[1] 郭雷[1] 赵天云[1] 钱晓亮[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710129

出　　处：《红外与激光工程》2012年第11期3102-3107,共6页Infrared and Laser Engineering

基　　金：航空科学基金(20080153002)

摘　　要：针对红外图像中的混合噪声,提出了一种自适应混合双边滤波算法。首先对双边滤波原理进行了分析,提出不能滤除强高斯噪声和脉冲噪声是由于双边滤波引入灰度域权值而带来的固有不足,因此根据双边滤波算法的特点设置了一种像素间的相似度,并以该相似度为基础将双边滤波不能滤除的强噪声点进行了标记,仅对红外图像中标记出的强噪声点进行中值滤波以减少图像模糊,对普通噪声点采用灰度方差自适应双边滤波以保留更多边缘特征。自适应混合双边滤波能够有效滤除红外图像中的高斯噪声、脉冲噪声以及由其组成的混合噪声,同时在滤波过程中并不降低双边滤波保留红外图像边缘特征的性能。仿真实验结果表明,与传统双边滤波、改进的双边滤波以及各项异性扩散-中值滤波算法相比,该算法无论是滤除红外图像的混合噪声还是保留边缘特征都较为优越。In view of the mixed noise in the infrared image, the adaptive hybrid bilateral filtering was proposed. The principle of bilateral filter was analyzed. It was put forward that the shortage of bilateral filtering which couldn＇t filter out strong Gauss noise and impulse noise was inherent. Therefore, the average similarity was defined for bilateral filtering. On the basis of the average similarity, noise points were marked with strong noise points which couldn＇t be filtered out by bilateral filtering and ordinary noise points which could be filtered out by bilateral filtering. To reduce image fuzzy, only the strong noise points were filtered by median filtering. To keep more edge character, ordinary noise points were filtered by adaptive bilateral filtering. Simulation results show that the hybrid bilateral filtering keeps more edge character, and filters out Gaussian noise and the mixed noise which consist of Gaussian noise and impulse noise effectively. The algorithm is superior to several other methods in subjective evaluation and objective evaluation.

关 键 词：混合双边滤波 平均相似度 中值滤波 混合噪声 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]