Reference-guided structure-aware deep sketch colorization for cartoons

Feature (linguistics) Sketch Color correction Color balance
DOI: 10.1007/s41095-021-0228-6 Publication Date: 2021-10-27T16:03:16Z
ABSTRACT
Abstract Digital cartoon production requires extensive manual labor to colorize sketches with visually pleasant color composition and shading. During colorization, the artist usually takes an existing image as guidance, particularly when colorizing related characters or animation sequence. Reference-guided colorization is more intuitive than other hints, such points scribbles, text-based hints. Unfortunately, reference-guided challenging since style of colorized should match reference in terms both global local In this paper, we propose a novel learning-based framework which colorizes sketch based on feature extracted from image. Our contains extractor extract image, network generate multi-scale output images by combining feature, discriminator improve reality Extensive qualitative quantitative evaluations show that our method outperforms methods, providing superior visual quality consistency task reference-based colorization.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (48)
CITATIONS (16)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....