數位相機彩色濾鏡陣列影像雜訊消除方法

Abstract

當今時下，數位相機已幾乎完全取代傳統底片相機，舊有的底片也有產品將之翻拍成數位照片，這個趨勢已如潮水般滾滾而來，不只是數位相機，舉凡能數位化的產品都逃不過數位化的命運，只因數位資訊便於處理、儲存以及傳遞。 數位相機中的影像處理是經過了很多道的程序而成，幾乎所有的程序都會增強雜訊的作用，所以降低雜訊的影響最好的辦法就是在任何影像處理的程序之前就先將影像來源本身具有的雜訊濾除掉，而影像雜訊濾除最困難的就是在濾除雜訊的同時也保存了影像的輪廓和高頻訊號。本篇論文提出了另一種消除影像雜訊的架構，由三個想法所組成，一個是在影像區塊中將正在處理的像素歸類給某一預設圖樣，既然可被歸類給某一圖樣，就可以對這圖樣中的元素平均以降低雜訊，另一個是獲取校正過後的相機雜訊特性，掃瞄出其雜訊特性並以標準差來表示，以便於用來判別正在處理的影像區塊是否可以判定為均勻影像，進而做出最強的濾除效果。最後一個是運用肉眼視覺系統不易察覺到那些藏在多紋路、高反差區域的影像雜訊，所以不去濾除那些不易被察覺的雜訊以保存更多的影像輪廓訊號。

Nowadays, traditional film camera has almost been replaced by digital camera in commercial market. This trend is not only on camera, but also on any product in which the signal can be digitalized, since digital information would be much convenience to be processed, stored, and transmitted. Image processing in digital camera consists of many processes. Almost all of the processes would enhance noise added in images. The best way to make noise strength be minimized is to reduce noise in front of any image processing. The difficulty of image denoising is always to preserve edge information, and filters out noise in flat area simultaneously. In this paper, we have presented a denoising method which consists of three ideas. One is to filter noisy pixel based on nearest pattern to keep edge information, another one is to use camera noise characteristic to judge the uniformity of current processed area and the last one is to make use of the property of spatial masking to keep edge information again on the highly texture area.