Abstract:

According to the low resolution characteristic of infrared video image,an improved super-resolution reconstruction algorithm of infrared video image is proposed in this paper.It researches the sparse representation for the low-resolution infrared video image patches,and gets the optimum coefficients of sparse representation by Least Angle Regression(LAR) algorithm.By reconstruction constraints,the final high-resolution video image can be obtained through the optimun coefficients.Experimental results show that,for infrared video images,the proposed algorithm can improve the resolution of the input image significantly and achieve better outcomes in visual effects and objective evaluation,compared with Bicubic,NE and UVII algorithms.

Key words: sparse representation, infrared video image, Least Angle Regression(LAR) algorithm, super-resolution

摘要：

针对红外视频图像低分辨率的特点,提出一种改进的红外视频图像超分辨重建算法。研究低分辨率红外视频图像块的稀疏表示,通过最小角度回归算法求解得到稀疏表示的最优化系数,并利用该系数产生重建约束后的高分辨率视频图像最终解,实现对红外视频图像的超分辨率重建。实验结果表明,对于红外视频序列图像,该算法在主观视觉效果和客观评价方面均优于Bicubic,NE和UVII算法。

关键词: 稀疏表示, 红外视频图像, 最小角度回归算法, 超分辨率

CLC Number: 

• TP391