基于色差空间的Bayer 图像的迭代插值算法

单CCD/CMOS 传感器相机捕捉图像信息靠在传感器表面覆盖一层颜色滤波阵列(CFA)，经过CFA 后每个像素点只能获得物理三基色(红，绿，蓝)其中一种分量。另外缺少的两种颜色分量，需要通过周围像素的值来估算。首先利用55 模板内的像素来估计插值的方向并用最优的权重系数来插值G 分量。其次利用了基于有理函数的二维插值算子在色差空间插值R(B)处缺少的B(R)分量。再次利用色差插值G 处缺少的R 和B 分量。最后，使用方差约束条件，迭代插值过程被重复多次直到达到了最优的插值结果。通过在24 幅柯达图片以及笔者相机拍摄的图片上的Matlab 仿真实验，结果显示，被提出的算法无论是在视觉方面还是在量化的数据方面都表现出了优势。

Iterative interpolation algorithm of Bayer images based on color difference space

Signal CCD/CMOS sensors capture image information by covering the sensor surface with a color filter array (CFA). For each pixel, only one of three primary colors(red, green and blue) can pass through the CFA. The other two missing color components are estimated by the values of surrounding pixels. The first step was to estimate interpolation direction taking advantage of the pixels in 55 template and use the optimal weighting factors to interpolate G components. The second step was to interpolate R(B) components at the location of B(R) using the interpolation operator based on two-dimensional rational function. The third step was to interpolate R and B components at the location of G components by color difference interpolation. Lastly, the iterative interpolation repeated until approaching the optimal results using variance-constrained condition. Through Matlab simulate experiments based on 24 Kodak images and the images captured from our camera, the proposed algorithm outperforms both in visual and numerical aspects.