Motion-guided semantic alignment for line art animation colorization

Ning Wang, Wei Wang, Hairui Yang, Hong Zhang, Zhiyong Wang, Zhihui Wang, Haojie Li

Abstract: Highlights•We propose MGC-Net for motion-guided colorization using pixel-to-pixel color mapping.•The HCD module ensures coarse pixel alignment via motion distillation loss.•The CR module refines alignments using a contextual dependencies.•Extensive experiments show MGC-Net outperforms current line art animation colorization methods.