图像抽象化的实时增强型绘制

Real-Time Enhancing Image Abstraction

现有的图像抽象化绘制算法主要采用基于图像分割或边缘保持的平滑滤波技术,导致所生成的抽象化图像经常存在明暗对比度较低、画质昏暗模糊、画面色彩表现力不强等问题,为此提出一种增强对比度的图像抽象化绘制算法.首先利用结构张量平滑技术生成光滑连续、显著特征保持的特征流场,然后基于该流场对图像进行沿边缘切向流方向的自适应平滑和沿梯度方向的对比度增强处理,最后使用改进的基于流场的双边滤波或者Kuwahara滤波算法实现图像抽象化的实时绘制.实验结果表明,该算法可显著改善已有抽象化方法的绘制质量,抽象化图像的边界特征明显增强,画质更清晰、层次感更分明.

To address the limitation that the result generated by existing methods looks somewhat washed out because of missing contrast, this paper presents an improved non-photorealistic rendering technique for abstracting images. A smooth and feature-preserving edge tangent flow field is firstly constructed from the eigenvectors of the smoothed structure tensor. Then an adaptive smoothing operation is performed along the stream lines defined by the edge tangent flow, and after that the difference-of-Gaussians-based unsharp masking is performed in the direction of gradient for enhancing local scene contrast. At last, an improved flow-based bilateral filter or a separable implementation of Kuwahara filter is adopted to achieve the enhanced abstracted images in real-time. Experimental results show that the proposed algorithm outperforms the existing methods in terms of feature preservation and visual appreciation.