【作者】 吴炜； 王美洁； 李智； 李康丽； 刘凯；

【Author】 WU Wei;WANG Mei-jie;LI Zhi;LI Kang-li;LIU Kai;School of Electronic and Information Engineering,Sichuan University;School of Electrical and Information Engineering,Sichuan University;

【机构】 四川大学电子信息学院； 四川大学电气信息学院；

【摘要】 利用可见光图像与红外图像之间存在的互补性和相关性,提出一种基于多传感器的红外图像超分辨方法。首先,利用自适应边缘检测算法提取红外图像边缘,并根据红外图像边缘与可见光图像对应区域的相关程度将红外图像边缘分为相关边缘和非相关边缘;然后,采用二次关系模型对相关边缘区域进行建模,通过该模型利用可见光信息估计红外图像的高频信息;最后,利用迭代反向投影法(IBP)对估计的高分辨率红外图像进行优化获得最终的高分辨率红外图像。实验结果表明,本文算法获得的小区图像,十字路口图像和道路图像的峰值信噪比(PSNR)分别比Choi算法高2.9dB,1.44dB和1.11dB。另外,利用本文算法复原的红外图像具有更好的视觉效果,更逼真、更接近于原始高分辨率图像,复原出的高分辨率红外图像无论在主观效果上还是在客观评价指标上都取得了较好的结果。更多还原

【Abstract】 On the basis of complementary and relevance between a visible image and an infraved image,a new method which is called multi-sensor super resolution method was proposed.Firstly,an adaptive edge detection algorithm was used to extract the edges of the infrared image,and edges were divided into related edges and nonrelated edges according to the relevance of the edges of infrared image and corresponding visible image area.Then,aquadratic relationship model was built to estimate the high frequency patch of infrared image in the related edge region.At last,Iterative Back Projection(IBP)method was used to optimize the estimated super-resolved image.Experimental results indicate that the Peak Signal to Noise Ratios(PSNR)of the community image,crossroad image and the road imagein this method are respectively 2.9dB,1.44 dB and 1.11 dB,higher than that of the Choi algorithm.Experimental results also show that the proposed method enables reconstructing images to be better visual results and closely resembling the original high resolution images.The reconstructed images achieve better results on subjective visual effects and objective assessments.更多还原