具有瑕疵移除能力的彩色濾波器陣列 插補演算法

Abstract

近年來，數位相機逐漸普及，也越來越受消費者青睞，普遍被當成日常生活中捕捉影像的一種方式。 基於節省成本和縮小尺寸上的考量，大多數的數位相機使用電荷耦合感光元件或者互補式金屬氧化半導體感光元件來捕捉現實場景中的彩色影像。 然而，單一 CCD 或 CMOS 感光元件無法提供足夠的色彩解析度。而大多數光學廠所採用的解決方案是在感光元件的上面覆蓋一層色彩濾光片，有如規律的馬賽克排列一般(例如：綠、紅、藍、綠)。這種濾光片就是所謂的色彩濾波器陣列。 因為真實場景的每個位置都是彩色的；然而，經過色彩濾波器陣列濾光之後，每個像素就只剩下濾波器相對應位置的顏色，丟失了三原色中的另外兩個顏色。因此，我們必須從適當鄰域中的像素點計算局部的紋理、色彩特性，進而重建出近似真實場景的彩色影像。上述這種從原始資料重建回彩色影像的過程就是所謂的色彩插補或者去馬賽克演算法。 我們基於殘餘色彩插補演算法的精神提出改良，在這篇論文中，提出一個色彩插補演算法 CFAIAAR，我們比較了五種不同的色彩插補演算法，我們發現，CFAIAAR 不僅保持更清晰的邊緣、細緻紋理細節，還能有效的移除假色、摩爾波紋、和迷宮紋理。

Recently, the digital still cameras (DSCs) are popular with consumers as a device to capture digital images easily. For cost and size consideration, most DSCs and industrial cameras use Charge-Coupled Device (CCD) or Complementary Metal-Oxide-Semiconductor (CMOS) sensor to capture the varied colors in a scene with a single sensor chip. However, the single CCD or CMOS sensor does not provide sufficient color resolutions. The solution for most DSC designers is to cover the sensor’s surface with a mosaic of colored filters: Color Filter Array (CFA). Since there is only one color array element in each pixel, the other two missing color elements must be estimated. Thus, the recovery of full color images from CFA-based sensors requires a process of estimating values of missing color elements at each pixel by its adjacent pixels. This reconstruction process of a full color image from a color mosaic is commonly called color interpolation or color de-mosaicking algorithm. In this thesis, a de-mosaicking technique is proposed to interpolate three color channels (Red, Green, and Blue) using Residual Interpolation and embedded artifact removal method based on median filter. We have compared this new approach to five different methods, and our method can not only outstandingly preserve more edge details and handle fine textures well but also remove most color alias, Moiré pattern, and maze pattern.