近年來,為了減少產品的體積及製作成本,大部分消費型數位相機都採用前方覆蓋一層彩色濾光陣列(CFA)的單一感光元件來擷取影像。由於CFA的結構,每一個像素都只能得到紅、藍、綠三原色其中一個顏色的資訊。色彩插補演算法主要的目的即是利用鄰近像素所得到的顏色資訊來估測CFA影像所失去的顏色。雖然目前許多色彩插補演算法已被提出,然而在計算複雜度與插補結果上無法同時得到令人滿意的結果。有鑑於此,本論文提出一個低計算複雜度且可以有效提升插補品質的演算法。此方法主要有四個特點,分別為,利用七個係數取代原有五個係數的方向低通濾波器來插補綠色,利用可變大小的方向濾波器來尋找影像邊緣的部分,利用比例式的權重插補綠色,以及使用改良式的中值濾波來改善插補後的影像。我們除了使用常見的Kodak 24張圖片之外,還使用其他兩組色彩飽和度較高的影像來進行分析和比較。實驗結果證實,本論文所提出的色彩插補演算法,在PSNR值或是SCIELab等客觀的評估中可以得到比近期所提出的演算法還要好的結果,而在主觀的視覺評估方面也可以得到不錯的視覺效果。
Recently, in order to reduce the product size and the manufacturing cost, most commercial digital cameras use a single sensor covered with a color filter array (CFA) to capture the scene. Based on the CFA structure, each sensor pixel only samples one of the three primary color values and the other two are missing. The purpose of demosaicking algorithm is to estimate the missing two color values by neighboring samples. Although many demosaicking algorithms have been proposed, they can not satisfy both the computational complexities and the quality of reconstructed images at the same time. Because of this reason, we propose a new algorithm which is low complexity and effective to improve image quality. This method has four characteristics which are seven-coefficient FIR filter, variable-size classifiers, proportional weighting sum of the candidates, and modified median filter refinement. In our experiment, we use not only the 24 Kodak testing images but also two sets of high saturated images to demonstrate the performance. Experimental results show that the proposed algorithm can obtain higher PSNR results and better visual quality than previous demosaicing techniques.