請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/6545
標題: | 自然影像的機率分析 A Probabilistic Analysis of Natural Images |
作者: | "Yang, Yao-Hsiang" 楊耀翔 |
指導教授: | 王傑智 |
共同指導教授: | 陳祝嵩 |
關鍵字: | 自然影像統計,降維,光照不變特徵抽取,貝氏分析,維納濾波, natural image statistics,dimension reduction,illumination-invariant feature extraction,Bayesian analysis,Wiener filtering, |
出版年 : | 2012 |
學位: | 碩士 |
摘要: | 自然影像統計,對於計算視覺社群的研究者而言,在過去二十年間早已引發相當的興趣。已有不少研究報告在許多的不同條件下收集到的影像資料之不同特徵。相較於其他不對影像作明確假設的應用,基於自然影像統計假設的許多應用研究也已證明相當有用並易於理解。
在本文中,我希望摘要那些過去研究的實驗結果與解釋。而後,我希望調查這些事實在研究中可以扮演的角色。在闡釋理論的架構之後,兩項新的應用,分別利用費雪判別分析與維納濾波,將被推演用以展示這一方法的效力。我希望這些初步的結果能顯示足夠的前途,讓人們對未來發展更加完備的影像分析之形式理論有足夠信心。 Natural image statistics has long been interested by the researchers of the computer vision community during the last twenty years. There has been reports on many distinct characteristics of image data collecting under various conditions. Several applications based on the assumptions derived from natural image statistics have been proved to be both useful and intelligible in compare to those works which do not employ any explicit assumptions about images. In this thesis, I wish to summarize those experimental results and explanations in previous studies. Then I wish to investigate the role of these facts in the research. After explicating the theoretical framework, two novel applications incorporating the Fisher's discriminant analysis and Wiener filtering are conduct in order to demostrate the power of this approach. I wish these rudimentary results are shown to be promising enough to make people more confident about the possible future toward a more complete formal theory of image analysis. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/6545 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 資訊網路與多媒體研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-101-1.pdf | 1.29 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。