Integration of Adaptive CFA Demosaicking with Bad Pixel Removal

In the manufacturing process of image sensors for digital cameras, some defective pixels known as bad pixels will inevitably be produced and more bad pixels may emerge with camera usage over time. These defective pixels can interfere with the demosaicking process, and in this paper, we propose an adaptive method to integrate bad pixel removal with the demosaicking process so that those defective pixels will be excluded in the interpolation process. Depending on the bad pixel location, our proposed adaptive method will adaptively change the order of interpolation so that the bad pixel will be outside the region of interpolation, i.e. if a bad pixel is closer to the point of interpolation, a lower order will be chosen. On the other hand, if the bad pixel is further away, a higher order will be selected for higher accuracy. It has been shown that our proposed adaptive CFA demosaicking method outperforms other existing techniques in terms of interpolation accuracy and bad pixel removal capability.