Image Demosaicing Based on Generative Adversarial Network

Figure 12

Comparison of reconstructed images from different demosaicing methods for the 1st image in the McMaster database. (a) Ground truth image; (b) enlarged portion in the black box; (c–l) the results by using Bilinear, TL, HQL, Zhang’s, LDI-NAT, ARI, MLRI, RI-GBTF, and DnCNN methods, as well as the proposed method, respectively; (m–v) the residual images corresponding to (c–l).