SIA OpenIR  > 光电信息技术研究室
单帧红外图像噪声去除方法研究
Alternative TitleResearch of Single-frame-based Noise Removal Method for Infrared Images
李方舟1,2
Department光电信息技术研究室
Thesis Advisor赵耀宏
Keyword非均匀性校正 条纹噪声 随机噪声 单帧图像 非局部均值
Pages79页
Degree Discipline模式识别与智能系统
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract本课题主要研究的是基于单帧图像的红外图像噪声的去除方法,目的是提高红外图像的信噪比,为红外图像的后续的检测,识别,跟踪等应用提供良好的基础。本文首先介绍了红外图像的条纹与随机噪声模型,并基于一组黑体实验图像对非制冷红外成像系统中温漂引起的条纹噪声的性质进行了分析,用垂直方向局部线性模型对含噪图像进行建模。在此基础上,用有色高斯模型描述图像块中条纹和随机混合噪声,并提出了一种基于图像梯度的条纹噪声强度估计方法。本文提出了一种基于加权最小二乘的红外图像条纹噪声去除方法,算法首先用水平方向一维加权最小二乘滤波对含噪图像进行滤波,作为对真实的图像信息的初步估计。然后用垂直方向局部加权岭回归对条纹校正系数进行回归,其中加权岭回归中的权值来自图像局部方差作为边缘检测算子,该方法能有效减少对条纹校正项的错误估计。针对传统的白噪声去除算法在处理条纹和随机混合噪声时产生的残留条纹和滤波效果下降等问题,本文对传统的非局部均值去噪算法进行了分析,并在此基础上提出了一种基于非局部均值的改进马氏距离的块相似性度量方法,并且度量方法中的距离函数能对图像局部复杂程度进行自适应调整。模拟含噪数据和真实数据实验结果均表明,本文单帧去条纹算法相比传统方法,能有效避免残留条纹现象的产生;相比于传统的单帧去白噪声方法,本文的改进非局部均值算法在图像复杂和平坦区域均能够较好的去除混合噪声。
Other AbstractThis topic mainly studies single-frame-based infrared image noise removal methods. The purpose is to improve the signal-to-noise ratio of infrared images, and provide a good foundation for the subsequent detection, recognition and tracking of infrared images. This paper first introduces the stripe and random noise model of infrared image, and analyzes the properties of stripe noise caused by temperature drift in uncooled infrared imaging system based on a set of blackbody experimental images. The vertical linear model is used to construct the noisy image observation model. On this basis, the colored Gaussian model is used to describe the stripe and random mixed noise in the image patches, and a stripe noise intensity estimation method based on image gradient is proposed. In this paper, a method of stripe noise removal based on weighted least squares is proposed. The algorithm firstly filters the noisy image with horizontal one-dimensional weighted least squares filtering as a preliminary estimate of real image information. Then, the stripe correction coefficients are regressed by the vertical direction local weighted ridge regression. The weight in the weighted ridge regression comes from the image local variance as the edge detection operator. This method can effectively reduce the erroneous estimation of the stripe correction term. Aiming at the problems of traditional white noise removal algorithm in dealing with stripe and random mixed noise, such as residual stirpes and filtering quality degradation, this paper analyzes the traditional non-local means denoising algorithm and proposes a improved algorithm. A block similarity measure method that improves the Mahalanobis distance is used, and the distance function in the measure method is adaptively adjusted to the local complexity of the image. The experimental results of simulated noisy data and real data showed that our single-frame-based destriping algorithm can effectively avoid the residual stripe phenomenon compared with the traditional method. Compared with the traditional single-frame-based white noise removal method, the improved non-local means algorithm can remove mixed noise better in both complex and flat regions.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25171
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
李方舟. 单帧红外图像噪声去除方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
单帧红外图像噪声去除方法研究.pdf(26070KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李方舟]'s Articles
Baidu academic
Similar articles in Baidu academic
[李方舟]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李方舟]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.