Cartoon image colorization based on emotion recognition and superpixel color resolution

With the development of artificial intelligence technology, it is possible to automatically colorize the hand drawn sketch by machine. Researchers have conducted in-depth and meticulous study on hand-drawn manuscript recognition, generation, and retrieval. For the emotion-based line art colorization, facial expression should be also extracted from the image itself. To solve the recognition and colorization problem, this paper has proposed an algorithm with the DenseNet network for emotional recognition of anime faces and performed a two-stage interactive coloring method in view of superpixel color analysis features. Among them, the superpixel color analysis technology used a simple linear iterative clustering of SLIC algorithm. According to the experiment, the prompt color information of corresponding position could be generated through the emotion recognition result. After the prediction of superpixel color analysis by GAN (generative adversarial network), the original cartoon image could be rendered with suitable color scheme. The visualization results proved that our algorithm proposed would effectively realize the emotion-based line art colorization of high interactivity and reasonable color distribution.