Saliency Detection-Based Mixture of Reality and Non-Photorealistic Rendering Effects for Artistic Visualization

摘要: In this paper, we bring the novel idea to automatically combine Non-Photorealistic Rendering (NPR) effects with real-world images based on saliency detection. Noticing that key of NPR is focus enabling a wide variety expressive styles for digital visual art (e.g. painting, sketch, and cartoon), such mixture Reality NRP always provides an extremely intriguing sense beyond original content. Technically, given input image, devote fast approach convert it into manga or pencil sketch on-the-fly. Moreover, guided by hierarchical detection strategy, can be finished in most effective way. On other hand, proposed 'RealMeetsArt' system also function let user manually select interested foreground regions. User easily fine-grained area only several stroke drawings.