A Study on Neural Style Transfer Methods for Images
Junzhe Liao · 2022
Image style transfer techniques have been around for almost twenty years. In particular, the rise of deep learning has gradually influenced traditional image transfer techniques, eventually giving rise to new neural image transfer techniques. This paper covers an introduction to traditional style transfer techniques, including Stroke-based rendering, image filtering, image analogy and texture synthesis, as well as a categorization of various current classic neural image style transfer methods including slow neural method, fast neural method and generative adversarial networks (GAN) based neural style transfer method. In addition, this paper covers commonly used training datasets for style transfer. Finally, the challenges of neural image style transfer are discussed, and the possible future research directions are contained.