Fundamental Study on Caricature Generation System from the Viewpoint of Kansei

Abstract

The present paper is aiming to build a system for generating facial caricatures that looked like hand-drawn ones by introducing the following techniques; "ruling exaggeration limit", "face parts", "simulating human touch". It was suggested that addition of kansei information like brush touch results in improvement of subjective impression. Further, We're attempting to implement "automatic feature point extraction by using AAM" for automatic caricatures generation from an input face image and "outputting caricature to SVG format (vector data)" for flexibility and extendibility, and the main parts of the system were almost completed.