利用双边滤波器和DoG算子的人像卡通化新算法

A New Approach for Cartoon-Like Stylization of Video Portrait Based on Bilateral Filter and DoG Operator

利用双边滤波器在平滑图像的同时又能保持边缘的良好特性,结合DoG(高斯差分)算子,提出人像卡通化的新算法.该算法首先应用双边滤波器对原始的彩色图像进行平滑,在获得颜色主要特征的同时,去除一些不重要的细节信息,简化了低对比度区域;然后,融入用DoG算子检测到的边缘曲线来增强高对比度区域;最后进行软量化以获得卡通图像.实验结果表明:相对经典的图像抽象化方法而言,该算法获得了更好的图像卡通化效果和更快的速度.

Using the good characteristic that bilateral filter can also preserve edges while smoothing image,combining with DoG(difference of Gaussians)operator,this paper proposes a new approach for cartoon-like stylization of video portrait.First,the approach applied bilateral filter to smooth the original color image.While getting the dominant features of color,it could removal some trivial details and simplify the low contrast regions.Then,the edge curves detected by DoG operator were integrated to enhance the high contrast regions.At last,the soft quantization was applied to stylize portrait.The experimental results show the approach achieves better effect of image cartoon-like stylization and faster speed relative to classic method of image abstraction.