Color Filter Array Demosaicking Algorithm Based on Convolutional Neural Network
Hsi-Hao Chang, Xin-Hong Lai, Junliang Chen, Ting‐Lan Lin, Chiung-An Chen, Shih‐Lun Chen · 2020
This paper presents a new demosaicking algorithm based on machine learning for color filter array (CFA) images. The proposed algorithm includes two convolutional neural network (CNN) machine learning methodologies. The first is Pooling which provides a method for judgment and the other is Shared Weights which reduces the parameter requirement and improves the overall performance of machine learning. Five types of patterns and eight interpolation methods of blocks are selected by the machine learning to improve the quality of demosaicking images. The experimental results show that the average PSNR can reach 35.05 dB and the average SSIM can reach 0.9848 by the proposed algorithm.