Novel color demosaicking for noisy color filter array data
Novel color demosaicking for noisy color filter array data
复制标题
DOI：
10.1016/j.sigpro.2011.08.009
复制
发表时间：
2012-02
期刊：
影响因子：
--
通讯作者：
Yu Zhang;Guangyi Wang;Jiangtao Xu;Zaifeng Shi;Dexing Dong
中科院分区：
文献类型：
--
作者：
Yu Zhang;Guangyi Wang;Jiangtao Xu;Zaifeng Shi;Dexing Dong
单传感器数字彩色静态/视频相机使用颜色去马赛克来从滤色器阵列（CFA）数据再现全色图像。由于图像采集过程中引入的传感器噪声，插值图像的质量会下降。许多传统的去马赛克-去噪解决方案采用通道相关噪声模型，其可能比信号相关噪声模型更不适合CMOS/CCD图像传感器。本文将小波子带分解与合成应用于信号相关噪声模型下的CFA数据插值。本文的主要工作包括：（1）将LMMSE与小波域的统计计算相结合，将信号相关噪声分为加性噪声和乘性噪声，对信号相关噪声进行抑制。(2)在CFA数据中，验证了当前像素点与相邻像素点之间的数量关系。仿真和真实的CFA图像被用来比较所提出的算法与文献中报道的最先进的技术。实验结果表明，我们的方法优于他们的去马赛克性能和计算成本，当他们处理噪声的彩色滤光片阵列数据。
Single sensor digital color still/video cameras use color demosaicking to reproduce full color images from color filter array (CFA) data. The quality of interpolated image will be degraded due to the sensor noise introduced during the image capture process. Many conventional demosaicking–denoising solutions adopt the channel-dependent noise model, which may fit the CMOS/CCD image sensor less than signal-dependent noise model. In this paper, the wavelet sub-band decomposition and synthesis are applied to interpolate the CFA data with signal-dependent noise model. The major contributions of this work include: (1) The combination of LMMSE and statistical calculation in wavelet domain are utilized to suppress the signal-dependent noise, which is separated into additive noise and multiplicative noise. (2) In CFA data, it has been verified that the quantitative relationship between the current pixel and the adjacent pixel, which locate in the same edge. Both simulated and real CFA images are employed to compare the proposed algorithm with the state-of-the-art techniques reported in the literature. The experimental results confirm that our method outperforms them both on demosaicking performance and on computational cost, when they process the noisy color filter array data.