Conclusion

Most of the novel demosaicing algorithms try to achieve a better approximation on edges in the CFA- based image. However, it is hard to find a balance between the performance and the computational complexity. The gradient method proposed in [8] returns an outstanding result but with a extremely high computational complexity cost[1]. Data-dependent triangulation and high quality linear interpolation are of resonable computational complexity, but their results are not as attracitve as [8].

There are also some works on natural image properties[9] and post-processing the recovered image[10]. Since it is hard to reconstruct the exact color information at the color-missing pixels, it may be a good compromise to reconstruct a image "making people feel good."

< Result > < Reference >
< Home >