Chinese Journal of Computers   Full Text
Title	A Demosaicing Algorithm Based on Bayer Pattern and Local Correlation Model
Authors	ZHOU Guo-Rui1) SUN Shi-Xin1) WANG Wen-Jiang2)
Address	1) (School of Computer Science and Engineering, University of Electronic Science and Technology, Chengdu 610054) 2) (Engineering Department, Longcheer Co.Ltd, Shanghai 200233)
Year	2007
Issue	No.7(1179—1185)
Abstract &   Background	Abstract First, a high-frequency similar model is presented based on the integration of color difference model and wavelet transform theory, and then a condition is deduced based on theoretical analyses. In order to construct the model, this condition should be met in the process of wavelet transform. A method for reasonably integrating the characters of Bayer pattern with the process of wavelet transform is presented. If use this method to reconstruct R and B planes, it can greatly decrease the reconstructing time by avoiding iteration. The ratio of computational cost used by this method and used by the one which proposed by Gunturk, et al. is 19∶432. Finally, a demosaicing algorithm is proposed. Compared with other typical algorithms, the above method shows better on both demosaiciing performance and computational cost. keywords Bayer pattern； color difference model； high frequency similar model； wavelet transform background A full-color signal is usually composed of three color planes and, accordingly, three separate sensors are required to record them respectively. To reduce the cost, many cameras use a single sensor covered with a color filter array (CFA). In the CFA-based sensor configuration, only one color is measured at each pixel and missed out two color values. Demosaicing refers to the problem of reconstructing the original full-resolution color signal from the downsampled vision. The objective of demosaicing is to reconstruct signal as close to the original as possible. Meanwhile, the computational complexity of demosaicing needs to be kept low for the reason of being cost effective in practical applications. Modeling the correlation among three color planes plays a critical role in demosaicing. Ignoring such interplane dependency often renders the demosaiced signal suffering from annoying artifacts caused by color misregistration. Meanwhile, it shows that the color correlation model significantly affect the computational cost of demoscaing in many literatures. This paper presents a high-frequency similar model based on integration of color difference model and wavelet transform theory. According to this model, a demosaicing algorithm is proposed, which can greatly decrease the computational cost by avoiding iteration and improve the quality of reconstructed signal.