Automatic hair parts segmentation for 2D animation production

Abstract

In the production of 2D animation based on illustrations, it is necessary to divide the target illustration into parts by category, such as hair, face, and body. Usually, the partitioning process is performed using graphics software, but when creating complex animations, the number of parts becomes large and time-consuming. In this paper, we propose a method to improve the segmentation accuracy of hair by automatically segmenting it into parts using U-net, which is one of the semantic segmentation methods, and by combining it with post-processing based on image processing.