基于图像变化检测的CFA插值算法

A Change Detection Based Demosaicking Approach

提出了基于变化检测的CFA插值算法。该算法分为两个步骤:首先采用矢量线性相关法对CFA图像的绿色分量图像进行变化检测,然后在发生变化的区域内进行插值,未发生变化区域内的图像数据被CFA插值后直接拷贝到当前帧的相应位置,避免了在每帧图像内逐点进行插值计算。与现有的插值算法相比,CFA插值算法复杂度大大降低,同时获得较高质量的视频图像。

In this paper, a change detection based demosaicking approach is introduced. The proposed demosaicking method consists of two successive steps： an image change detection step that indicates changed parts between two consecutive green component images by linear depend- ence detector for vectors, and an interpolation step applied in changed area. Contrastly, the full color image data in unchanged area is directly copied from previous interpolated image. Experimental results indicate that the proposed framework exhibits excellent performance in terms of the commonly used objective criteria and at the same time its computational complexity is lower than other analogous methods.