Research On Cartoon Colorization And Generation Based On Style

In this paper, we mainly discuss two aspects: two-dimensional cartoon and style.Two-dimensional cartoon animation is a forms of art that many people like verymuch. Each cartoon shows its comprehensive characteristics, namely”style”, which isunique. For example, motion, scene, expression, color, and so on. It is a new directionfor research about reusing the style to produce new cartoons. The main contents ofthis paper are showed below:1.We extract color from source cartoon using histogram, and then, we analyze thecolor characteristics of the cartoon.2.We extract the relationship between light and shade of image pixel, viaimage multiscale decomposition technique.3.We use optimization of colorization method to colorize images, with light and shaderelations as the basis of color diffusion.4.We compute the termination conditions of color diffusion via Canny and Gaborwavelet.On the other hand, we propose cartoon retargeting technique. Main contents are:1. Cartoon motion capture. A background subtraction based method ofextracting pose sequence of cartoon character is designed. And then, wepropose a method of simplifying binary image of contours of cartooncharacter.2. Cartoon retargeting. We propose a new bounded biharmonic weight deformationalgorithm, which can ensure reasonable deformation.This paper analyses style of two-dimensional cartoon via color characteristics andmotion characteristics, and how to transfer the style. At the same time, it sets up anoverall framework that including deformation, image colorization, imagesegmentation and other technology, this framework is useful and significant.