Semantic-Sparse Colorization Network for Deep Exemplar-Based Colorization

Exemplar-based colorization approaches rely on reference image to provide plausible colors for target gray-scale image. The key and difficulty of exemplar-based is establish an accurate correspondence between these two images. Previous have attempted construct such a but are faced with obstacles. First, using luminance channel the calculation inaccurate. Second, dense they built introduces wrong matching results increases computation burden. To address problems, we propose Semantic-Sparse Colorization Network (SSCN) transfer both global style detailed semantic-related in coarse-to-fine manner. Our network can perfectly balance local while alleviating ambiguous problem. Experiments show that our method outperforms existing methods quantitative qualitative evaluation achieves state-of-the-art performance.