研究テーマ : パーツ単位のモーフィングにもとづく似顔絵生成

研究概要 :


Theme : Face Cartoon Generation based on Parts-by-parts Morphing

Abstract :

It is a difficult problem to generate a cartoon that captures the characteristic of the face from the limited number of sample cartoons. We focus on a face cartoon generation technique that does not have any limitation of the number of samples. The technique first decomposes an input image into part images including eye, brow, mouth and nose. The technique then calculates similarity between each of the input part images and prepared sample part images, and selects several similar sample part images. The technique synthesizes the part cartoons by morphing the sample part cartoons corresponding to the selected sample part images. Final cartoons are generated by combining the morphed cartoons. The technique also represents expressions on the cartoons by transforming the each of the part cartoons. Our system can represent the own features of faces. We also expect the system enables smooth communications on the social networking services.