红外序列图像处理关键算法研究

发布时间：2018-10-23 07:40

【摘要】：红外成像技术又称作热成像技术,是夜视技术的一种,这里特指被动式红外热成像技术,它是一种通过将物体的热辐射转换为电信号,从而获取人眼可以识别的可见光图像的成像技术。红外热成像技术不受天气、光照、电磁干扰等影响,是一种可以全天时,全天候工作的成像技术。经过多年的大力发展,红外热成像技术在军事、民用等许多领域都得到了广泛的应用,但仍然存在以下两个方面的问题：(1)受生产工艺、探测器材料以及图像处理技术的限制,红外图像的空间分辨率和温度分辨率都不高。而且红外图像是黑白图像,不具备颜色信息,限制了图像携带的信息量,增加了图像解读的难度；(2)红外成像设备被大量应用于导弹、飞行器、舰艇、单兵等运动平台上,平台抖动也会极大的影响成像效果。因此,从红外热成像技术的理论出发,以改善现有问题为目的,促进红外热成像技术在军用和民用领域的应用,对于加快我国国防建设步伐,提高国民经济各领域的技术水平,有着非常重要的意义。深入研究了红外成像设备应用于运动平台时面临的实际问题,以提高红外序列图像质量为目标,分别提出了新的图像增强算法、稳像算法和彩色化算法,具体内容如下：(1)红外图像增强技术。从红外成像特点出发,深入研究了红外图像增强面临的关键问题,提出了两种改善红外图像质量的图像增强算法,分别是基于自适应尺度因子的多尺度视网膜皮层(Multi-scale Retinex,MSR)算法和基于双边滤波器(Bilateral Filter, BF)和MS R的自适应红外增强算法。对MSR算法进行改进,重新定义MSR算法中的关键尺度因子,提出了基于自适应尺度因子的MSR算法。算法根据图像细节自适应的调整MSR算法中最为关键的尺度因子,改善了MSR算法对红外图像的增强效果,并通过实验证明了本算法的有效性。提出了一种基于BF和MSR的自适应红外图像增强算法。通过自适应的选择BF中的关键参数,是一个新的具有图像增强和动态范围压缩的自适应增强算法。最后通过实验证明了本算法的有效性、实用性和适应性。(2)红外序列图像稳像技术。从基于运动平台的红外成像设备的成像特点出发,分析了视频序列抖动的原因及特点。讨论了热像仪成像的几何模型以及图像的运动模型,提出了一种新的运动估计算法并对运动滤波算法进行改进,提出一种新的稳像算法框架。引入了BRISK(Binary Robust Invariant Scalable Keypoints)算法,提出了基于BRISK算法的运动估计算法,可以更有效的实现全局运动估计。实验证明了所提出算法的有效性和优越性。通过对几种常用的运动滤波方式的分析和讨论,引入了基于Kalman粒子滤波的运动滤波算法。该方法能够更好的实现复杂扫描运动和随机振动的分离。并提出了更为高效的边界区域补偿办法,以有效实现图像的全景输出。进一步通过稳像实验,结合上述关键算法,对实地采集的视频序列进行了稳像实验,并使用不同的算法对实验结果进行了比较和验证,实验结果证明了本文所提出算法的可行性和有效性。(3)红外图像的彩色化技术。讨论各种典型彩色化技术,重点研究基于色彩传递的彩色化算法。详细分析了影响色彩传递算法的若干因素,深入研究了色彩空间对彩色化效果的影响,在此基础上对传统算法提出了改进,使其更适用于红外图像的彩色化处理。将局部线性嵌入(Locally Linear Embedding, LLE)算法中的降维思想应用到色彩传递算法中,提出了一种新的基于LLE的色彩传递算法,并通过实验验证了其有效性。
[Abstract]:Infrared imaging technology, also known as thermal imaging technology, is a kind of night vision technology. It is a passive infrared thermal imaging technology, which is an imaging technique for obtaining visible light image recognizable by human eye by converting thermal radiation of object into electric signal. Infrared thermal imaging technology is not affected by weather, illumination, electromagnetic interference and so on. It is an imaging technology that can work all-day and all-weather. After many years of vigorous development, the infrared thermal imaging technology has been widely used in many fields, such as military and civil, but there are still two problems: (1) the limitation of production process, detector material and image processing technology; Neither the spatial resolution nor the temperature resolution of the infrared image is high. moreover, the infrared image is a black-and-white image, does not have color information, limits the amount of information carried by the image, increases the difficulty of image interpretation, and (2) the infrared imaging device is widely applied to a moving platform such as a missile, an aircraft, a ship, a navigation system and the like, The platform jitter also greatly affects the imaging effect. Therefore, from the theory of infrared thermal imaging technology, in order to improve the existing problem, it is very important to promote the application of infrared thermal imaging technology in military and civil fields. In order to improve the image quality of infrared sequence, a new image enhancement algorithm, image stabilization algorithm and colorization algorithm are presented, which are as follows: (1) Infrared image enhancement technology. Based on the characteristics of infrared imaging, the key problem of infrared image enhancement is studied deeply. Two image enhancement algorithms for improving the quality of infrared image are proposed, which are multi-scale retinal cortex based on adaptive scale factor (Multi-dimension Retinex). MSR) algorithm and adaptive infrared enhancement algorithm based on bilateral filter (BF) and MS R. The MSR algorithm is improved, the key scaling factor in MSR algorithm is redefined, and MSR algorithm based on adaptive scale factor is proposed. The algorithm improves the enhancement effect of MSR algorithm on infrared image, and proves the effectiveness of the algorithm. An adaptive infrared image enhancement algorithm based on BF and MSR is proposed. It is a new adaptive enhancement algorithm with image enhancement and dynamic range compression by choosing the key parameters in the adaptive selection BF. Finally, the validity, practicability and adaptability of the algorithm are proved by experiments. (2) Infrared sequence image stabilization technology. Based on the imaging characteristics of the infrared imaging equipment based on the motion platform, the causes and characteristics of video sequence jitter are analyzed. In this paper, the geometric model and motion model of image are discussed, a new motion estimation algorithm is proposed and the motion filtering algorithm is improved, and a new stable image algorithm framework is proposed. The BRISK algorithm is introduced, and a motion estimation algorithm based on BRIK algorithm is proposed to realize global motion estimation more effectively. The experiment proves the validity and superiority of the proposed algorithm. Through the analysis and discussion of several commonly used motion filtering methods, a motion filtering algorithm based on Kalman particle filtering is introduced. The method can realize the separation of complex scanning motion and random vibration better. A more efficient method of boundary region compensation is proposed in order to effectively realize the panoramic output of the image. Based on the key algorithm described above, the experiment results are compared and verified by using different algorithms. The experimental results show the feasibility and validity of the proposed algorithm. (3) colorization technology of infrared image. In this paper, a variety of typical colorization techniques are discussed in this paper, and the colorization algorithm based on color transfer is studied. In this paper, some factors which influence the color transfer algorithm are analyzed in detail. The influence of color space on colorization effect is deeply studied. On the basis of this, the traditional algorithm is improved so that it is more suitable for colorization processing of infrared image. In this paper, a new LLE-based color transfer algorithm is proposed, and its validity is verified by experiments.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】