基于原对偶算法的全变差图像复原

来源 :昆明理工大学学报(自然科学版) | 被引量 : 1次 | 上传用户:mzt1989
下载到本地 , 更方便阅读
声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架
论文部分内容阅读
图像在采集、存储、处理和传输过程中,由于成像系统、存储设备、传输介质和处理方法的不完善,导致图像质量的下降,这种现象叫做图像退化.图像复原就是对退化的图像进行处理,尽可能恢复出原始图像的真实信息.由Rudin等提出的ROF模型能很好的保持图像的边缘,并广泛应用于图像复原问题中.本文正是在此模型的基础上修改其正则项,建立一新的模型,然后利用原对偶算法对新模型求解并证明了该算法的收敛性.实验结果表明,新模型在信噪比上有很大的改进,提高了图像的质量且抑制了阶梯效应.
其他文献
This article summarizes a joint research projec undertaken under the Risk Management Solutions, Inc(RMS) banner to investigate some of the possible approaches for agricultural risk modeling in China.
期刊
The Tohoku Regional Bureau(TRB) of the Ministry of Land, Infrastructure, Transport and Tourism(MLIT) performed various actions in response to the 2011 Great East Japan Earthquake and Tsunami. The tota
期刊
This article addresses three large earthquake disasters in Iran: Tabas in 1978, Rudbar in 1990, and Bam in 2003. Lessons and "Lessons Learned" from these three earthquake disasters were investigated t
期刊
We present preliminary results from the experimental investigation of the response of the atmosphere due to the impact of powerful shock waves.The response is evidenced as ultra low frequency electrom
期刊
Technology innovations are important for firms to compete in the market, whereas spillovers are unavoidable. The research issue of this article is how imitations impact on R&D behaviors. This article
期刊
This paper proposes a kind of facial expression recognition technology based on the D-S theory with weighting coefficient of information fusion, in the background of Affective Computing theory, moreov
期刊
This article presents an efficient and robust algorithm that compute the intersection curve of a ruling surface with a plane. These ruling surface is represented by S(s,t)=B(t)+sV(t) where B(t) is the
期刊
In order to deal with contradiction between big dynamic range and single means to acquiring binarized vertical edge image in texture-based adapting license plate location(TALPL),research on acquiring
期刊
为解决多人行为识别中高维特征空间、角色分配不准确和复杂的时间结构等问题,本文分析了多人行为的特点,提出了一种递归的多层随机网络模型。该模型通过多层网络表达行为的多尺度特性,并由高层体现行为的长时间依赖性。通过对观测的分解大大降低了特征空间的维数,从而降低了问题的复杂度,并在一定程度上消除了目标角色分配不准确带来的影响。实验结果表明本文提出的模型比其他常用模型具有更好的识别效果,即使对复杂行为依然具
期刊
基于面向特定应用的可配置处理器架构及其设计方法,设计并完成了一种视频媒体编解码片上系统芯片,它具有通用数字信号处理器的柔性编程及特定目标应用时的高性能等特点。该视频编解码片上系统由编码和解码2部分组成,编码和解码部分都采用相同的媒体信号处理架构。媒体信号处理编码、解码架构中分别包含一个8发射超长指令字数字信号处理器核,还包括实现视频媒体应用的专用数据传输单元,变长编解码单元以及接口单元,可以完成H
期刊