- 學位論文
影像脈衝雜訊移除方法之研究與濾波器設計
Image filter design for impulse noise revomal
摘要
影像雜訊移除在影像處理中是一個很熱門的領域，在很多系統中，影像資料常常因為感測器的產生過程或是在不穩定的傳輸通道中受雜訊干擾，使得部份像素的值會和原始像素不同，因此雜訊的移除是必須的。在本篇論文中我們把焦點放在移除影像中的脈衝雜訊消除。然而，影像雜訊移除在雜訊密度較高時，像素周圍也是雜訊點的機率也相對提高，增加了雜訊移除的困難度與效能的迅速降低。 在本篇論文中，我們利用了鄰近像素的方向性likelihood判斷中心像素的周圍像素是否遭受干擾，進行Error Excluded的動作。目的是為了先將掃描視窗內的雜訊先行移除，將周圍的雜訊像素移除後，再取中值運算得到Error Excluded Median，另外，我們也檢測中心像素和Error Excluded 後的像素的方向性likelihood的關係。最後利用這兩項的檢測結果，各取兩檢測的優點，避免雜訊移除的同時，誤將未受干擾的像素判定為雜訊。最後由實驗結果可以看出我們的濾波器能有效地在高雜訊時移除雜訊，並且不會破壞影像的細節。
並列摘要
Images are often corrupted by impulse noise due to error generated in noisy sensors or communications. It is essential to remove the impulse noise before some image processing, such edge detection, image compression etc. In this thesis, we will focus on the impulse noise revomal effectively in digital image. However, when the images are highly corrupted, a large number of impulse pixels may connect into noise blotch. The filtering performance of filters will degrade rapidly. To solve this problem, we utilize the directional likelihood to detect whether the neighbor pixel of center pixel is noisy pixel or not. Excluding the noisy pixel from the vector of median operation thus we obtain the Error Excluded Median. On the other hand, we can acquire the error excluded directional likelihhood. By combinning the two testing results, we judge whether the center pixel is noisy pixel or clean pixel. Finally, experiment result indicates that our proposed scheme outperforms the convention scheme.
參考文獻
延伸閱讀
- 吳振豪（2007）。適用於廣域脈波雜訊移除之脈波雜訊偵測與濾波〔碩士論文，大同大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0081-0607200917242047
- 謝政修（2015）。高隨機脈衝雜訊影像之模糊象限分群濾波器〔碩士論文，國立宜蘭大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0046-0110201512393700
- 游宜樺（2014）。脈衝雜訊影像之高效疊代中值濾波器〔碩士論文，國立宜蘭大學〕。華藝線上圖書館。https://doi.org/10.6820/NIU.2014.00203
- 侯理中（2015）。有效移除影像中脈衝雜訊之演算法〔碩士論文，國立宜蘭大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0046-0608201513201200
- 曾哲威（2022）。Self-Supervised Deep-Learning Based Image Filter for Impulse Noises〔碩士論文，國立宜蘭大學〕。華藝線上圖書館。https://doi.org/10.6820/niu202200056