一种基于小波变换的多尺度铅笔画生成算法

An pencil drawing generation algorithm based on wavelet transform multiscale

近年来,非真实感绘制技术(non-photorealistic rendering,即NPR)在娱乐视频、医学等方面都有广泛的应用。现存的许多自然图像铅笔画生成算法,效果都不太理想,存在如铅笔画边缘线条粗重或者层次不够突出、缺少细节等一系列问题。在分析了传统的铅笔画生成算法的基础上,提出了一种改进的自然图像的铅笔画生成算法。该算法仿照艺术家手工绘制铅笔画的过程,使用小波变换的多尺度方法生成铅笔线条轮廓图,使得到的线条轮廓图保留了更多的细节信息且边缘线条不会十分粗重;再利用直方图匹配的方法得到与艺术家手工绘制铅笔画相同的色调,并使用一个优化将色调与铅笔画的纹理图结合从而得到色调纹理图;最后将线条轮廓图与色调纹理图合并得到效果良好的铅笔画。通过将该算法应用在各类图像上,发现由该算法得到的铅笔画效果与现存铅笔画生成算法相比,不仅保留了更多的细节,而且更加接近艺术家手工绘制的铅笔画。

In recent years, non-photorealistic rendering （NPR） technology is widely used in the entertainment video, medicine, etc. The generation algorithm of pencil drawing results of many natural image are not satisfactory, and a series of problems exist, for example, the pencil edge lines are so coarse, the painting level is not prominent or the details are not enough. We mainly study the improved pencil drawing generation algorithm. This paper analyzes the traditional pencil drawing generation algorithm, and compare it with some pencil drawing generation algorithm, on this basis, an improved pencil drawing generation algorithm of the natural image is put forward. This algorithm is generated by modeling after the artist % hand drawn pencil drawing process, the line profile map is extracted based on wavelet transform multiscale method to make the line profile map retain more details and the edge lines will not be very coarse; then, using the histogram matching method to obtain the tone which is the same as the tone of artist manual pencil drawing, and the color texture map is obtained using optimization method by combining the tone and pencil texture map; finally, combining the line profile map and the tone texture map to obtain a pencil drawing with good result. Applying the algorithm in this article on all kinds of images, the results show that this algorithm retain more details and the results are more similar to those of the artist＇s hand drawn pencil drawing compared with the existing pencil drawing generation algorithm.