空中视频序列的分辨率增强技术的研究

【作者】 郭锦杰；

【导师】 娄震；

【作者基本信息】 南京理工大学 ， 模式识别与智能系统， 2008， 硕士

【摘要】 本文主要针对基于空中视频图象序列的图象分辨率增强技术进行研究,该技术主要是通过获取得到的低分辨率视频图象序列应用超分辨率重建算法来重建高分辨率的图象以达到图象分辨率增强的目的。通过超分辨率重构算法重建高分辨率图象,既提高了分辨率,又提高了系统的效能,是现有既定条件下比较可行的一种图象分辨率增强方法。本文就基于空中视频图象序列的超分辨率重构技术开展了一系列理论研究和实验探索,主要做了以下几个方面的工作。首先,介绍并解释了与该课题相关主要术语,介绍了该课题的应用背景以及国内外研究现状。以期使读者对该课题从宏观上有一个粗略的了解。建立一个正确的图象退化模型是图象复原和图象超分辨率重建的前提,本文比较详细的介绍了图象退化模型建立的过程,构造更高分辨率的图象就是图象的上抽样过程,离不开对图象的插值放大处理,因此本文简单介绍了一些基本的图象插值算法。基于本文所研究问题的应用环境,占有很大一部分比例的视频图象是运动退化的本文详细研究并实现了频域和空域中的运动退化图象复原算法。并取得了较为满意的恢复结果。另外,本文简要的介绍了当前比较流行运动估计方法,光流场方法和模块匹配方法。本文研究并实现了基于凸集投影(POCS)和迭代反卷积(IBP)两种超分辨率重建算法来重建更高分辨率的图象以达到图象分辨率增强的目的。本文详细分析了这两种算法各自的优缺点,分析得知,由于其良好的抗噪能力和能充分利用先验信息的能力,使得POCS算法比较适合本文所研究问题的应用环境。最后对本课题做了总结和展望,提出了后续可以进行的一些工作。更多还原

【Abstract】 The research work focuses on the resolution enhancement by the video sequences in the air, which is a technique of estimating a high-resolution from video sequences to enhance the resolution of image. This technique reconstructs the image by algorithm of high-resolution. It raises the resolution as while as the efficiency, so that it has become one of the most available methods to enhance resolution of image.The thesis summarized the academic and experimental research on the super-resolution reconstruction by the air video sequence, and the detailed research works can be described as following:At the very beginning, we introduced the main gloss correlate to the problem. And then we took the pictures of the resolution testing board and other practicalities to make the reader have a macroscopical view about the problem.The precondition of the image restoration and the super-resolution is building a correct model of degradations. The thesis introduced the process of building degradation model. Since super-resolution reconstruction is the process of up-sampling the low-resolution images, we introduced some basic interpolation algorithm for digital images.The thesis introduced the process of building degradation model. Since super-resolution reconstruction is the process of up-sampling the low-resolution images, we introduced some basic interpolation algorithm for digital images. Besides that, to evaluate the quality of processed image, we introduced some evaluation functions to evaluate image quality.In this right condition, the motion blurred images hold a large proportion in the video sequences, so one of our main tasks is to restore these motion blurred images. The thesis research and implement the main restoration algorithms, including frequency domain and special domain processing and obtain approving results. Besides that, we introduced the popular algorithm for motion estimation.The thesis researched and implemented the projection onto convex sets super-resolution reconstruction algorithm and the iterative back projection super-resolution reconstruction algorithm to enhance the resolution of the image. We analysis the advantages and the disadvantages of the two algorithms .And then we got the conclusion, for it’s noise resistibility and priori information capability, POCS algorithm is fit for the right problem which the thesis researched. In the end of the thesis, we prospected the future of the technology and analyzed the necessary development in this field.更多还原