摘要超分辨率重建是一种由一序列低分辨率退化图像重建一幅(或序列)高分辨率清晰图像的第二代复原技术,可以有效地弥补从硬件上提高图像分辨率的不足,在航空成像、遥感成像、医学成像等众多领域具有广泛的应用前景。60137
本文主要研究成果有如下三点:
(1)在简要说明了噪声和模糊核的概念后介绍了一种噪声图像中高斯噪声方差的盲测定方法和一种利用局部图像结构方向导数的模糊核多尺度迭代估计方法。并且描述了这两种方法的具体步骤和流程图。
(2)构建了超分辨复原模型和图像的退化过程,并在此基础上介绍了基于水平和垂直梯度的L1稀疏性图像先验模型,为之后实现重建算法提供了理论依据。
(3)通过Matlab,本文给出了基于分层贝叶斯的多幅图像超分辨率的模块化设计,实现了基于分层贝叶斯的超分辨重建软件,该软件包括非盲超分辨、盲超分辨和盲参数估计等功能。此软件操作简单,运行良好,使用大量的测试用例表明了该软件具有很高的精确性。
毕业论文关键词: 超分辨率重建,噪声,模糊核,分层贝叶斯,软件设计
毕业设计说明书(论文)外文摘要
Title Based on hierarchical Bayesian multiple images super resolution algorithms and software
Abstract
Super-resolution reconstruction is the second generation of recovery technology to use a sequence of degenerative low resolution images to rebuild one (or sequence) high resolution clear image. It can effectively make up for the inadequacy of improving the image resolution from the hardware. It has a broad application prospect in the aviation imaging, remote sensing imaging, medical imaging and many other fields.
This paper has the following three main research results:
(1)After briefly explaining the concept of noise and blur kernel, this paper introduces a blind determination method of a noise image’s Gaussian noise variance and a multiscale iterative method using local image structure directional derivative to estimate the blur kernel. And this paper describes the specific steps and flow chart of the two methods.
(2)Building a super-resolution restoration model and image degradation processes, and on this basis, introducing a based on the horizontal and vertical gradient L1 sparse image prior model. It provides a theoretical basis for implementing reconstruction algorithm in the future.
(3)By Matlab, this paper gives an image super-resolution modular design based on hierarchical Bayesian. And implementing a super-resolution reconstruction software based on hierarchical Bayesian. This software includes non-blind super resolution, blind super resolution and blind parameter estimation and other functions. This software is simple, running well, using a large number of test cases show that the software has a very high accuracy.
Keywords: Super-resolution reconstruction, noise, blur kernel, Hierarchical Bayesian, software design.
1 绪论 5
1.1 课题背景与意义 5
1.2 超分辨模型与算法的研究现状 6
1.3 本文工作 7
2 图像超分辨问题中的退化参数估计方法 8
2.1 引言 8
2.2 单幅图像的噪声方差估计方法