一种新的红外图像细节增强算法  

Novel algorithm of detail enhancement for IR image

作　　者：杨家红[1] 向锦[1] 李翠红[1] 危德益[1] 

机构地区：[1]湖南师范大学工学院,湖南长沙410006

出　　处：《激光与红外》2012年第5期579-583,共5页Laser & Infrared

基　　金：湖南省自然科学基金重点项目(No.08JJ3131);湖南省教育厅重点项目(No.10A076)资助

摘　　要：红外成像系统在数字化时使用14位(或16位)数据来量化一个像素,具有动态范围大、含噪多和对比度低等特点,因此在进一步处理之前往往需要进行必要的增强。目前常用的红外数据增强都是基于普通方法压缩至8位后完成,许多细节信息已在压缩过程中损失,因而增强效果有限。本文把增强过程放在从14位到8位的压缩过程中,提出一种双域滤波及改进的最值归一化对比度增强算法的红外图像增强方法。首先使用双域滤波将原始14位图像的细节部分与基本部分分开,压缩基本部分的动态范围、保留或增强细节部分,然后使用改进的最值归一化方法增强图像对比度,最终得到可以在普通显示器上处理的8位图像数据。该方法在压缩红外图像数据宽度的同时保留了细节信息,取得了良好的实验结果。Infrared images are usually digitized to 14 bits（or 16 bits）.The images often contain much noise and have low contrast,so they need to be enhanced before further processing.The common enhancement methods are usually carried out after the IR images have been compressed from 14 bit to 8 bit,which results in detail information lost.We here propose a new algorithm,by which the enhancement process is carried out along with the compressing procedure.The algorithm combines the bilateral filtering with the contrast strengthening by improved extreme value normalization.Firstly,we split the detail part from the basic part of the image,saving and enhancing the detail part while compressing the dynamic range of the basic part.Then we use the improved extreme normalization method to enhance the image contrast and get the enhanced 8 bit image data which can be displayed on normal monitor.The experiment results show that this method can make good compression while preserve the detail image information.

关 键 词：红外图像 双域滤波 最值归一化 图像细节增强 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]