基于视觉显著区域的红外图像增强算法研究  

Research on Infrared Image Enhancement Algorithm Based on Visual Saliency

作　　者：汪忱 张宝辉[2] 王虹[1] 李中文 

机构地区：[1]南京理工大学电子工程与光电技术学院,江苏南京210094 [2]北方夜视科技集团有限公司南京研发中心,江苏南京211106

出　　处：《红外技术》2017年第9期835-840,共6页Infrared Technology

摘　　要：人眼对于感兴趣区域的视觉关注度更高,为了实现对相关区域的图像增强,本文提出了基于视觉显著区域的红外图像增强算法。本算法通过去雾预处理使图像在检测时获取更加完整的显著性区域,再对该区域采用分层差分表达理论,把高层的差分向量换算到低层,最终计算显著性区域的差分向量以及图像的全局变换函数,实现图像增强。实验结果表明,本文算法对显著性区域的增强效果明显,合理地突出了显著性区域,增强图像的同时对原图边缘细节的保持能力很好。The human eye has a higher degree of visual acuity for regions of interest. In order to enhance such regions in images, this paper proposes an algorithm of infrared image enhancement based on visual saliency. First, the de-fog algorithm detects a more complete salient region in the image. Then, the theory of layered difference representation is applied to the region, and the difference vector of the upper layer is applied o the lower layer. Finally, the vector of the salient region and the global transform function of the image are calculated to obtain the enhanced image. The experimental results show that the proposed method not only highlights the salient region reasonably well, but also enhances it. The method also maintains the boundary of the original image during enhancement.

关 键 词：视觉显著性 分层差分表达理论 红外图像增强 去雾 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]