红外图像非均匀性校正算法研究

发布时间：2018-04-11 01:09

本文选题：红外成像技术 + 非均匀性校正　；参考：《南京理工大学》2017年硕士论文

【摘要】：近年来,红外成像技术不断进步,核心探测器不断更新换代,探测系统日益完善,尤其是反应灵敏、集成度高、耗能小的红外焦平面阵列的出现,使得红外成像技术在国家安防、环境监测、国民经济发展等各大领域发挥着不可替代的作用。然而红外图像上的条纹、畸变,使得探测器的温度灵敏度降低,探测距离缩短,严重影响了图像的质量,我们称这些固定的噪声为红外图像的非均匀性[Non-Uniformity,NU]。本论文以得到高分辨率红外图像为目标,首先对红外成像技术、红外探测器的发展及应用进行了阐述,紧接着对非均匀性在图像上的存在形式、产生原因和学术界普遍认同的定义进行了说明,指出非均匀性校正过程的机理。最后论文研究了定标类算法与两类基于场景的典型算法。一是统计类,典型算法有:时域高通滤波法、神经网络法、统计恒定法。另外是配准类,典型算法有:运动补偿、代数法和帧间配准。通过对比说明了各自的优缺点。为了克服传统的定标类算法需要黑体参考源和严格的线性要求等的不足,本论文提出基于焦平面均值响应的非均匀性校正算法。通过实验与对异常响应像元的校正,验证了新算法有着更好的线性性能,并且对二阶函数图像的校正效果明显优于传统定标法中的二阶拟合,非均匀性校正能力很强;为了克服传统的配准类算法不易配准,计算量大,硬件难以实现等的不足,本论文提出基于递归的矩阵校正法。并且通过实验评估,验证了新算法计算量较小,对存储空间的需求低,可以自适应调整参数,收敛性与稳定性都很好,校正后的图像可视性高的优势。
[Abstract]:In recent years, the infrared imaging technology has been improved, the core detectors have been updated and the detection system has become more and more perfect, especially the infrared focal plane arrays, which are sensitive in response, high in integration and low in energy consumption.Infrared imaging technology plays an irreplaceable role in national security, environmental monitoring, national economic development and other fields.However, the temperature sensitivity and detection range of the detector are reduced because of the fringe and distortion on the infrared image, which seriously affects the image quality. We call these fixed noises Non-UniformityNU in infrared image.This paper aims at obtaining high resolution infrared image. Firstly, the infrared imaging technology, the development and application of infrared detector are described, and then the existence of non-uniformity in the image is discussed.The causes and the generally agreed definitions are explained, and the mechanism of the nonuniformity correction process is pointed out.Finally, the scaling class algorithm and two typical scene-based algorithms are studied.One is statistical, typical algorithms are: time domain high pass filter, neural network, statistical constant method.There are also registration classes, such as motion compensation, algebraic method and frame registration.The merits and demerits of each are explained by comparison.In order to overcome the shortcomings of traditional calibration algorithms, such as blackbody reference sources and strict linear requirements, a nonuniformity correction algorithm based on focal plane mean response is proposed in this paper.Through experiments and correction of abnormal response pixels, it is proved that the new algorithm has better linear performance, and the correction effect of the second-order function image is obviously better than the second-order fitting in the traditional calibration method, and the ability of non-uniformity correction is very strong.In order to overcome the disadvantages of traditional registration algorithms, such as difficult registration, large computation and hard hardware implementation, this paper proposes a recursive matrix correction method.Through the experimental evaluation, it is proved that the new algorithm has the advantages of low computational complexity, low demand for storage space, adaptive adjustment of parameters, good convergence and stability, and high visibility of the corrected image.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】