EMD算法与图像时频分析
文献类型：学位论文
| 作者 | 崔峰 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2005-04-01 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 彭思龙 |
| 关键词 | 时频分析 EMD 四元数谱 纹理分析 |
| 其他题名 | EMD Algorithm and Image Time-Frequency Analysis |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 图像信号的时频分析是图像处理的重要领域和难点，其目的是分析图像的各种频率成份在空间上如何变化，利用信号局部的频率等特征来完成图像分析与处理的目的。本论文对图像时频分析的工具与算法进行了研究。 本论文重点对 EMD 分解和四元数谱分析等新颖的时频分析工具进行了深入研究，并在图像分析方面获得了实际进展。首先，根据一维 EMD 分解的特点提出二维可分离的 EMD 分解，将 EMD 分解这种一维信号时频分析新工具推广到二维信号做出了探索；然后，为了分析二维信号的解析信号，对近年来出现的四元数解析信号进行了研究，提出了四元数快速 Fourier 变换和反变换算法，并针对现有四元数解析信号相位定义的缺陷提出一种新颖的四元数解析信号相位定义。 本论文在图像时频分析的应用方面也进行了探索。首先，结合二维可分离EMD 分解和四元数解析信号的特点，提出了一种利用二维图像二维解析特征实现纹理图像分割的方法；然后，在二维可分离 EMD 分解的基础上，提出了一种新颖的纹理压缩方法；最后，在四元数 Fourier 谱快速计算算法的基础上，对四元数域纹理信号的 Wold 分解进行了分析。 理论分析与实验结果表明，本论文的算法在创新性、精确度、复杂度、适用性等方面有着各自的优势。为 EMD 分解和四元数谱分析的理论与在图像分析领域的应用作出了有益的探索。 |
| 英文摘要 | The time-frequency analysis for image signal is one of the significant and difficult domains in image processing technology. The aim of it is to analysis how the frequency components vary in local time or local space and utilize the time-frequency features, such as instantaneous frequency, to accomplish image processing assignments. The tools and algorithms for image time-frequency analysis are just researched in this thesis. This thesis researches EMD and quaternionic spectrum analysis deeply, which are novel time-frequency analysis tools, and achieves actual evolve in image analysis. At first, the 2-D separable EMD based on the characters of 1-D EMD is introduced in this thesis, which extends the 1-D EMD to 2-D; Then, in order to analysis analytic signal for 2-D signals, this thesis researches the quaternionic analytic signal and introduces a fast quaternionic Fourier transform and inverse quaternionic Fourier transform algorithm. Simultaneously, this thesis puts forward a novel phase definition for quaternionic analytic signal, which overcomes some shortcomings of the exist definition. In the application domain of image time-frequency analysis, this thesis also does some research. First, an unsupervised texture segmentation algorithm is proposed in this thesis, based on the characters of 2-D separable EMD and quaternionic analytic signal; Second, this thesis introduces a novel texture compression method based on 2-D separable EMD; At last, it analysis texture Wold decomposition in quaternionic spectrum domain, based on fast quaternionic spectrum computing algorithm. Theory analysis and experiment results indicate that the algorithms proposed in this thesis have respective advantages, such as innovation, accuracy, simplicity, applicability, etc. This thesis explores a reasonable way to use EMD and quaternionic spectrum analysis in image analysis domain. |
| 语种 | 中文 |
| 其他标识符 | 200218014603197 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/5841]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 崔峰. EMD算法与图像时频分析[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2005. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。