兰州大学机构库 >数学与统计学院
基于展开网络的图像去马赛克方法
Alternative TitleAn Unfolding Network Method Based For Image Demosaicing
吴丰江
Subtype硕士
Thesis Advisor黄玉梅
2023-05-25
Degree Grantor兰州大学
Place of Conferral兰州
Degree Name理学硕士
Degree Discipline计算数学
Keyword图像去马赛克 image demosaicing Majorization-Minimization方法 majorization-minimization method 展开网络 unfolding network 优化模型 optimization model
Abstract

在拍摄彩色图像时, 彩色图像记录设备中的色彩滤波阵列一般记录到的是图像中每个像素的单通道像素值. 图像去马赛克问题是指利用单通道像素值, 构造有效方法恢复出每个像素另外两个缺失的通道像素值信息, 从而得到彩色图像的过程. 彩色图像在工农业生产和人们的实际生活中有着广泛的应用, 构造有效的图像去马赛克算法具有重要的理论意义和实用价值.

本文基于最大后验概率估计方法得到一个图像去马赛克优化模型, 基于求解优化模型的Majorization-Minimization方法构建了一个展开网络, 设计了展开网络的网络架构和损失函数. 对所构建的展开网络, 采用400对马赛克图像/彩色图像进行了训练. 证明 了展开网络所实现的图像去马赛克算法在一定条件下的收敛性. 数值实验中展开网络的测试结果表明, 本论文所提出的方法能够有效实现图像的去马赛克.

Other Abstract

When capturing color images, the color filter array in the color image recording device generally records the single channel pixel values of each pixel in the image. The problem of image demosaicing refers to the process of using a single channel pixel value to construct an effective method to recover the information of the other two missing channel pixel values for each pixel, thereby obtaining a color image. Color images have a wide range of applications in industrial and agricultural production and people' s actual lives. Constructing the effective image demosaicing algorithm has important theoretical significance and practical value.

In this paper, an image demosaicing optimization model is obtained based on the maximum posterior probability probability estimation method. An unfolding network is constructed based on the majorization-minimization method for solving the optimization model. The network architecture and loss function of the unfolding network are designed. The constructed unfolding network was trained using 400 pairs of mosaic/color images. Proved the convergence of the image demosaicing algorithm implemented by the unfolded network under certain conditions. The test results of the unfolding network in numerical experiments indicate that the method proposed in this paper can effectively achieve image demosaicing.

MOST Discipline Catalogue理学 - 数学 - 计算数学
URL查看原文
Language中文
Other Code262010_220200931731
Document Type学位论文
Identifierhttps://ir.lzu.edu.cn/handle/262010/534792
Collection数学与统计学院
Affiliation
兰州大学数学与统计学院
Recommended Citation
GB/T 7714
吴丰江. 基于展开网络的图像去马赛克方法[D]. 兰州. 兰州大学,2023.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Altmetrics Score
Google Scholar
Similar articles in Google Scholar
[吴丰江]'s Articles
Baidu academic
Similar articles in Baidu academic
[吴丰江]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[吴丰江]'s Articles
Terms of Use
No data!
Social Bookmark/Share
No comment.
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.