【作者】 庞博；

【导师】 高隽； 张旭东；

【作者基本信息】 合肥工业大学 ， 信号与信息处理， 2005， 硕士

【摘要】 超分辨率技术(Super Resolution SR),也称为High Resolution(HR),是图像处理领域的一个重要技术。普通的图像重建技术如去卷积技术只能将物体的频率恢复到相应的截止频率处,这样截止频率之外的能量和信息就不得不丢失了。超分辨率技术就是试图将截止频率之外的信息恢复,使图像尽量恢复场景真实性的技术。超分辨率技术已经在近年来国际上图像复原领域成为最活跃的研究课题。超分辨率技术(可以看成是一个视觉的逆过程)特指由一个序列低分辨率噪声图像来估计一幅或是一系列较高分辨率的原图像,同时还要能消除噪声以及光学元件产生的模糊。 在超分辨率图像重建的过程中,插值方法是很重要的一个因素,它可以消除采样误差以及虚假响应,但是插值可能会以系统性能损耗和增加计算复杂性为代价,而且不同的插值方法会使图像重建最终的结果有很大的差异。 本文围绕频域超分辨率图像重建计算模型展开,主要研究内容包括: 针对传统图像插值技术以及新兴图像插值技术分析了图像插值技术的分类以及不同方法的特点,并分析了各自的优缺点及其适用场合,从图像采样的角度分析了图像插值的必要性和重要性。论文通过传统方法和一些改进方法的比较,并通过实验说明了插值技术在图像处理中的作用。 研究并分析了国内外最新的超分辨率技术现状,针对当前图像成像系统分析了超分辨率技术的必要性和实用性,在总结和整理的基础上,针对目前超分辨率图像重建技术的技术分类和特点,重点分析了频域方法与空域方法的优缺点以及在理论和实际结合的基础上阐述了目前该技术的不足以及发展。 针对图像插值技术与图像超分辨率重建技术的结合,在运动信息估计计算模型的基础上,利用多帧低分辨率序列图像重建实验说明了插值技术在图像超分辨率重建中的重要性。更多还原

【Abstract】 SRC Super Resolution),which has an another name HR(High Resolution), is an important technique in image processing.The idea behind super-resolution imaging is to combine the information from a set of slightly different low resolution images of the same scene and use it to construct a higher resolution image. The differences between the low resolution images can have many origins, such as camera motion or change of focus.Some regular image reconstruction methods can reconstruct a high resolution image only within its cut-off frequency and have to lose the information beyond the frequency of the image. SR is the technique trying to avoid losing any of the information of image. It is one of the most active fields in the world of image processing recently. SR (which can be seen a inverse process of optic) reconstruct super (high) resolution image of one or set by using low resolution images (which have strong relationship between each other) without noise and aliasing.Interpolation is very important in the process of SR image reconstruction because of reducing noise and aliasing, but different methods also can bring cost of system computation and different affections to the result.In this paper we introduce interpolation and SR image reconstruction based on a model of one frequency SR reconstruction method.We first analysis the categories of interpolation and the difference between traditional and new methods. Interpolation is very important from the angle of image sampling. The affection of interpolation is showed by experiments.Secondly, we illustrate the art of state of super-resolution in the world and analysis of practicality of super-resolution. Based on the characteristic, the good and shortcomings of SR is analyzed.At last, we use a frequency domain technique to precisely register a set of aliased images, based on their low-frequency, aliasing-free part. After registration, with all the information of the set of low resolution, a super resolution image is then reconstructed using interpolation which is second phase of SR image reconstruction through values of different points in the low resolution images. Then we analyze affections of different interpolations to the reconstructed image.更多还原