Transferring Characteristic Proportions to Modify the Artistic Style of Cartoons

Transferring Characteristic Proportions to Modify the Artistic Style of Cartoons

Philip Buchanan
Lund University

Michael Doggett
Lund University

Computer Graphics International 2012, Bournemouth, UK, June

Abstract

This paper presents a new automatic algorithm for extracting properties of an individual artist’s style and applying those properties to other drawings. The algorithm can be used to ensure consistent artistic style across a project by automatically transferring a selected style to hand drawn images. We present an algorithm that uses multiple stages; subdivision of an image into selected geons, extraction of corresponding datasets from additional images, analysis and characterisation of the artist’s style and finally transferring the artist’s style to target images. When transferring the artist’s style we use a complexity sensitive scaling algorithm to avoid artefacts. We apply this algorithm to line drawn cartoons and demonstrate that artistic style can be transferred between cartoon objects. Our results show that analysis of object proportions can be used practically to ensure consistency for objects with the same Geon structure but drawn by different artists.

Downloads

Author generated version of the paper

Supplementary Material