绘制内容指导的铅笔画风格实现

发布时间：2018-05-27 20:40

本文选题：非真实感绘制 + 铅笔画风格　；参考：《计算机辅助设计与图形学学报》2017年07期

【摘要】：为了让铅笔画纹理的色调更接近画家的绘制效果,提出一种内容指导的铅笔画色调生成算法,以改进铅笔画纹理的生成效果.首先对大量绘制内容不同的铅笔画分类后进行色调直方图统计,拟合出适合每种类别的直方图曲线,并对输入图像用与之同类别的直方图曲线进行色调匹配,得到对应的铅笔画色调图;然后用与输入图像大小相同的随机噪声图为基底,对噪声图的每个像素进行不同次数的迭代相乘后产生与色调图匹配且分布一致的噪声场;最后对输入图像进行区域分割,在与分割区域对应的噪声场中用LIC算法生成卷积长度及卷积方向不同的铅笔纹理图,实现分区域不同方向的铅笔纹理的绘制.实验结果表明,由于采用内容指导的分类绘制,该算法生成的铅笔画在色调上更加贴近真实画家绘制的铅笔画,且纹理形式多样,整体效果较好.
[Abstract]:In order to make the tone of the pencil-drawing texture more close to the painter's rendering effect, a content-directed algorithm for generating the pencil-drawing hue is proposed to improve the effect of the pencil-drawing texture generation. First of all, a large number of pencils with different contents are classified, and then the histogram curve suitable for each category is fitted, and the input image is tonal matching with the histogram curve of the same category, and the histogram curve of the input image is matched with the histogram curve of the same category. The corresponding pencil-stroke hue map is obtained, and then the random noise image of the same size as the input image is used as the base, and each pixel of the noise image is iteratively multiplied with different times to produce a noise field matching and distributing the tone image. Finally, the input image is segmented, and the LIC algorithm is used to generate the pencil texture image with different convolution length and convolution direction in the noise field corresponding to the segmentation region. The experimental results show that the proposed algorithm is more close to that of the real painters in hue because of the classification and rendering with content guidance, and the texture forms are various, and the overall effect is better.
【作者单位】：云南大学信息学院;
【基金】：国家自然科学基金项目(61163019,61271361,61462093) 云南大学中青年骨干教师项目(XT412003) 云南省科技厅项目(2014FA021,2014FB113)
【分类号】：TP391.41

【参考文献】