摘要
现今已有大量的CFA插值算法问世,然而大多数只是针对独立使用或者是后期处理的场合。因为这些方法往往需要复杂的运算过程甚至迭代计算,所以不适合实时硬件实现。针对高清Bayer图像,提出了一种面向硬件的实时、高效的插值算法。该算法专注于插值过程的计算复杂度最小化以适合低资源利用量的实时硬件实现并且使最终输出的彩色图像保持较高的质量。使用加权梯度方法对每个像素的3乘5邻域分析得到最平滑的两个方向,丢失的两个颜色分量则沿该两个方向进行插值。采用了峰值信噪比(PSNR)和CIELAB颜色空间误差两种评判标准将该算法与目前一些常用的插值算法进行了比较。
There are currently a lot of methods available for CFA demosaicing,however most of them are designed for off-line or post-processing purpose,requiring complex computing process or even iterations which are not suitable for real time hardware implementation.In this paper,a cost effective method for high definition Bayer pattern image demosaicking dedicated to compact hardware implementation is proposed.In order to reduce the time and resources required for hardware implementation,the proposed method focus on minimizing the computational load of the demosaicing process while preserving high image quality.A weighted gradient approach is adopted to the 3 by 5 neighborhood around each pixel to determine the two smoothest directions,along which the two missing colors are then interpolated.A comparison between our method and a commonly used demosaicking method is given in metrics like peak signal to noise ratio（PSNR） and CIELAB color difference.
出处
《仪器仪表用户》
2011年第5期67-71,共5页
Instrumentation
