基于POCS的红外弱小目标超分辨率复原算法研究

	基于POCS的红外弱小目标超分辨率复原算法研究
	陈健
学位类型	博士
导师	高慧斌
	2014-07
学位授予单位	中国科学院大学
学位专业	机械电子工程
摘要	随着红外成像相关产业的兴起，红外成像技术具有的隐蔽性好、探测范围广、定位精度高、穿透距离远，以及轻质小巧、低耗可靠等优点备受青睐，已成为当前智能化光电探测发展的主流方向。然而，红外弱小目标的图像细节特征少、信噪比低等特点成为红外图像应用的瓶颈，如何提高红外弱小目标成像效果成为目前的研究热点。本文以“复原为本”为研究着眼点，利用超分辨率复原相关理论和技术，研究红外弱小目标超分辨率复原的理论和技术。本文主要围绕基于POCS的红外弱小目标超分辨率复原算法展开研究。针对红外弱小目标超分辨率复原中出现的问题，对传统POCS超分辨率复原算法进行了优化，提出了四种改进算法，提高了复原算法的性能，使其达到实时或接近实时，进而可以在实际红外图像处理系统中应用。本文提出了四种改进的POCS算法和一种新的超分辨率复原评价方法，并分别通过基于红外动态场景仿真系统实验和基于红外图像采集及处理系统实验，验证了改进算法和评价方法的有效性。本文的主要工作及创新之处在于：（1）针对传统POCS复原方法对噪声比较敏感的问题，将目前去噪效果较好的BM3D滤波方法和POCS复原方法相结合，对BM3D方法进行优化，提出了使用图像块的均值预筛选和限制分组图像块数目的方法，降低了BM3D方法的运算量。实验表明基于BM3D的POCS超分辨率复原算法能够在低分辨率图像包含噪声时，取得比传统POCS方法更好的复原效果，复原的高分辨率图像主观基本上看不出噪声。（2）针对传统的超分辨率复原评价体系只关注图像某一方面统计特性的缺点，提出了基于SSIM_NCCDFT的超分辨率复原评价方法。该评价方法结合了空间域的灰度均值、对比度以及频域自相关，能够同时评价超分辨率复原结果在空间域的复原效果和对频域信息的复原精度，实验表明该评价方法能够很好的评价超分辨率复原的结果，对超分辨率评价方法具有一定的指导意义。（3）针对POCS超分辨率复原算法迭代时间较长，无法满足光电探测系统实时性的缺点，提出了基于梯度图的快速POCS超分辨率复原算法。该算法根据图像的梯度分布对图像中的像素点进行分类，采用不同的迭代系数进行计算。在梯度越大的点，迭代步长越大；在梯度越小的点，迭代步长越小。改进算法能够较好的保留边缘信息并抑制噪声，进而在保证超分辨率复原性能的基础上大大缩短了运算时间。同时，提出了另外一种改进算法基于区域选择的快速POCS超分辨率复原算法。光电探测系统中我们关注的重点是目标区域，而这一区域通常只占很少的像素位置，因此通过阈值分割和合并找到所有目标区域并集，然后仅在这个目标区域并集上进行超分辨率复原。这样，去除了复原背景的巨大运算量，大大缩短了运算时间，使其达到实时或接近实时，进而可以在实际红外图像处理系统中应用。
其他摘要	With the spring up of the infrared imaging related industry, the infrared imaging technology has become the mainstream development direction of the intelligent photoelectrical detection, owing to good concealment, wide detection range, high positioning accuracy, long distant penetration, light weight, little volume, low power dissipation and high solidity. However, the features of the infrared dim-small target such as less details and low SNR become the bottleneck of the application of the infrared image. How to enhance the imaging effect of the infrared dim-small target becomes the hotspot of the research. Starting from the point of “restoration as foundation”, explores the theory and technology of the infrared dim-small target super-resolution restoration by utilizing the theory and technology of the super-resolution restoration. This paper mainly focuses on the research of the super-resolution restoration arithmetic of the infrared dim-small target based on POCS. For the super-resolution restoration problem of the infrared dim-small target, this paper optimizes the super-resolution restoration arithmetic of the traditional POCS. This paper proposes four improved arithmetic and improves the performance of the restoration arithmetic. Improved arithmetic can achieve real-time or near real-time. Improved arithmetic can be used in the actual infrared image processing system. This paper proposes four improved arithmetic and a new evaluation method of the super-resolution restoration. Then we verify the effectiveness of the improved arithmetic and evaluation method, by using the infrared dynamic scene simulation system and the infrared image processing system. The main work and innovation of this paper is: (1) For the noise sensitive problem of the traditional POCS restoration arithmetic, this paper combines the better de-noising BM3D filtering method and the POCS restoration arithmetic. We optimize the BM3D method. Furthermore we propose the method of mean pre-screened image block and limiting the number of packet image blocks, to reduce the computation of BM3D method. The experiments show that the POCS super-resolution restoration arithmetic based on BM3D can achieve better restoration effect than the traditional POCS method when the low resolution image contains noise, furthermore we can’t see noise on the high resolution image. (2) For the disadvantage of the traditional super-resolution restoration evaluation system only concerning about a particular aspect of the statistical properties of the image, this paper proposes the super-resolution restoration evaluation method based on SSIM_NCCDFT. This evaluation method combines the gray value and contrast of the spatial domain and the autocorrelation of frequency domain. So this evaluation method can evaluate the results of the super-resolution restoration in both spatial domain and frequency domain. Furthermore this evaluation method has some significance for super-resolution restoration evaluation. (3) For the long iteration of the POCS super-resolution restoration arithmetic and the shortcomings of incapability to meet optical detection system real-time detecting, this paper proposes a fast POCS super-resolution restoration arithmetic based on the grads graph. This arithmetic classifies image pixel according to the grads of the image, and then uses different iteration factor to calculate. The iteration step is bigger when the gradient is bigger and the iteration step is smaller when the gradient is smaller. This improved arithmetic can preferably retention edge information and suppression noise. Furthermore this improved arithmetic can guarantee the performance of the super-resolution restoration and greatly reduce the operation time. Simultaneously, this paper proposes another fast POCS super-resolution restoration arithmetic based on region selection. The optical detection system importantly focuses on the target area. The target area usually occupies a small pixel location. So we use Threshold segmentation and combination to acquire the union of all targets area. Then we execute super-resolution restoration only in the union of all targets area. In this way we decrease the huge computation of background restoration and greatly reduce the operation time. Furthermore the operation time achieve real-time or near real-time. So this super-resolution restoration arithmetic can use in the actual infrared image processing system.
语种	中文
文献类型	学位论文
条目标识符	http://ir.ciomp.ac.cn/handle/181722/41396
专题	中科院长春光机所知识产出
推荐引用方式 GB/T 7714	陈健. 基于POCS的红外弱小目标超分辨率复原算法研究[D]. 中国科学院大学,2014.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
陈健.pdf（8454KB）			开放获取	CC BY-NC-SA	请求全文