基于模糊算子的红外图像去模糊研究  

Research on infrared image deblurring based on blur operator

作　　者：张锦航 孙立辉[1] 姜军强 ZHANG Jin-hang;SUN Li-hui;JIANG Jun-qiang(Hebei University of Economics and Trade,College of Information Technology,Shijiazhuang 050061,China;Department of Mechanical Engineering,Xi′an University of Science and Technology,Xi′an 710054,China)

机构地区：[1]河北经贸大学信息技术学院,河北石家庄050061 [2]西安科技大学机械工程系,陕西西安710054

出　　处：《激光与红外》2023年第1期130-136,共7页Laser & Infrared

基　　金：河北省重点研发计划项目(No.20350801D)资助。

摘　　要：针对镜头抖动,目标移动等因素引起的图像运动模糊问题,本文提出了一种基于模糊算子的红外图像去模糊算法,使用深度自编码网络对数据集中的模糊算子进行编码,通过编码后的模糊算子去逼近一个未知的模糊算子并搜索对应的清晰图像,从而实现真实场景下红外图像去模糊,弥补了现有基于深度学习的图像去模糊模型在跨域应用时对真实场景下运动模糊图像去模糊效果较差的不足。在红外图像上的实验结果表明,相比于其他去模糊算法,本文提出的去模糊算法取得了更高的性能指标,恢复出的图像有着清晰的边缘轮廓和局部细节,显著提升了红外图像的清晰度。Aiming at the image motion blur caused by lens jitter, target movement and other factors, an infrared image deblurring algorithm based on fuzzy operator is proposed in this paper. The deep self-coding network is used to encode the fuzzy operator in the dataset, the encoded fuzzy operator is used to approximate unknown fuzzy operator and search for the corresponding clear image, thusthe infrared image deblurring in the real scene is realized, which makes up for the deficiencies of the existing image deblurring model based on deep learning in the cross-domain application of the motion blurred image in the real scene. The experimental results on infrared blurred images show that compared with other deblurring algorithms, the proposed deblurring algorithm achieves higher performance, and the recovered images have clear edges and local details, significantly improving the clarity of infrared images.

关 键 词：红外图像 深度学习 模糊算子 图像去模糊 自编码 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]