一种模糊红外图像人眼快速定位方法  

Fast Eye Detection from Blurred Infrared Images

作　　者：郑爽[1] 付冬梅[1] 

机构地区：[1]北京科技大学自动化学院,北京100083

出　　处：《红外技术》2016年第10期845-849,共5页Infrared Technology

基　　金：国家自然科学基金(61272358)

摘　　要：热像仪拍摄的红外图像能够反映人体表面的温度分布,进而间接反映其生理状态。但红外图像目标边界不清晰,与可见光图像相比目标更加模糊,因此对目标的定位和检测更加困难。本文研究并提出了一种快速自动定位人眼的方法,该方法从模板匹配的角度出发,结合模板和图像子区域的灰度相似度和纹理相似度进行匹配,同时利用Zernike矩和连通域分析筛选搜索点,减少了搜索次数。试验结果表明,本文提出的算法同经典的基于灰度相关的模板匹配算法相比速度更快,准确率更高。Thermal image taken by infrared thermography can reflect distributing of skin temperature and indirectly reflect the physiological status, but the boundary of the object in infrared image is not that clear. Compared with visible image, ‘the infrared one＇ is much more blurred, which makes it harder to locate and detect targets. In this paper, a novel method based on template matching is proposed to detect eyes quickly. It can locate the eyes correctly by means of gray scale correlation and texture correlation between template image and sub-images. And the method of Zernike moment operator and connected component analysis are adopted to reduce the search times. The experiments show that this algorithm performs better in detecting rate and accuracy than classical template matching algorithm based on gray correlation.

关 键 词：红外图像 人眼定位 模板匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]