Multi-GANs for colorize images with harmony state

Multi-GANs by inspiring of traditional GANs, divides each problem space into several smaller and more homogeneous subspaces. Multiple generative adversarial networks work together to achieve the highest output quality as a unite architecture. This paper presents colorization that based on two novelties including the cluster numbers and the color harmonies by Advanced Multi-GANs architecture. Advanced Multi-GANs can decide unsupervised about the number of clusters using the input test image. Its scene complexity that leads to much more realistic colorization. Also, we proposed color harmony to keep the harmony of the colors among a sequence of frames in video colorizing, which defines a rational relation between frames pixels and their generated colors. For avoiding changing colors of the same objects between video frames, we benefit the Color harmony. Evaluation our approach is formed with several protocols, including image and video colorization. In addition to visual qualitative evaluation, the efficiency of the proposed method is quantitatively measured in Advanced Multi-GAN framework. The comparison of our experimental results and the traditional approaches and state-of-the-art show much more realistic outputs.