【作者】 刘彬；

【导师】 刘秀平；

【作者基本信息】 大连理工大学 ， 计算数学， 2016， 硕士

【摘要】 随着科学技术的高速发展,人们对现代生活质量的需求越来越高。图像作为一种形象易懂的媒介信息,越来越受到人们的喜爱和关注。然而,在图像采集过程中,成像设备和场景之间不可避免地会出现一定程度的相对运动,导致图像质量下降出现模糊。基于此人们对图像去模糊进行了深入的研究,促使了视频监控,公共安全,医学处理及遥感卫星等领域的快速发展。尽管适宜光照条件下图像去模糊工作已经有了很大的进展,但是由于人们对夜晚缤纷亮丽梦幻般色彩的追逐,低光照下的图像去模糊工作才刚刚兴起。不同于适宜的光照,对低光照条件下的图像进行去模糊是一件困难的事情,由于场景的高对比度导致这些图像没有包含太多有用的结构信息,而这些信息对于模糊核的估计是至关重要的。本文主要关注的是低光照条件下单幅图像的盲去运动模糊问题。首先,我们阐述了图像去模糊问题的研究背景以及相关意义,并指出了现有方法的优势以及不足之处。其次,我们简要介绍了图像去模糊问题的相关理论基础并分析了在图像去模糊问题中常用的基于正则化的最大后验估计方法。最后基于夜景中特有的光条纹,我们提出了一种有效的图像去模糊算法。简要概述就是低光照图像中经常会含有一些明亮的光条纹,这对于模糊核的估计提供了很多有利的信息。一方面,对于求解一个非凸的去模糊优化问题,它们能够提供一个很好的初值；另一方面,它们记录了模糊核的路径。因此,基于光条纹我们提出了一种新的用于估计模糊核的先验。同时为了保证模糊核的形状和光条纹的形状相似,我们也给出了一种提炼光条纹形状的方法。最后在光条纹的帮助下,我们估计模糊核的过程不再需要去模糊方法中经常使用的启发式多级金字塔策略。大量的实验结果证明了我们提出方法的有效性。除此之外,我们把这种策略应用到现有的去模糊方法中会获得更好的实验效果。更多还原

【Abstract】 With the rapid development of science and technology, people begin to pursue higher quality of life. Images, as media information which are easy to understanding, become more and more popular and attract more and more attention. However, there inevitably be some degree of relation motion between the imaging equipment and the scene when taking photos, which leads to the deterioration of image quality. Based on these cues, people have made a deep research on image deblurring, which has prompted the rapid development of video surveillance, public safety, medical imaging and remote sensing satellite and other related fields.Although image deblurring has made great progress in suitable illumination conditions, it is just emerging in low-illumination conditions. Images that captured from low-illumination conditions is a challenging task, because these images contain few useful structures for kernel estimation. This paper mainly focuses on a single image blind motion deblurring problem. Firstly, we elaborate the research background and application value of image deblurring, and point out the advantages and disadvantages of existing methods. Secondly, we briefly introduce the relevant theoretical basis of image deblurring and analyze the regularization methods which are used in maximum a posteriori framework. Finally, based on the unique light streak in the night image, we propose an effective image deblurring algorithm.Images captured from low light conditions usually contain some light streaks, which are beneficial for estimating the blur kernel. One of our key observations is that these light streaks can provide a good initial value for a non-convex problem in kernel estimation. Another one is that they can record the track of the blur kernel at the moment when images are taken. Therefore, we propose a new prior for kernel estimation based on light streaks in this paper. Moreover, in order to ensure the shape of the blur kernel to be similar to that of light streaks during the updating, a new method is proposed to refine the shape of light streaks. With the help of the refined shape, our kernel estimation process does not require heuristic coarse-to-fine strategy, which is widely used in image deblurring methods. Quantitative experimental results show the effectiveness of the proposed method. In addition, we also demonstrate that the proposed method can be applied to existing deblurring methods to achieve better performance.更多还原