基于梯度修正和多向纹理提取的素描画生成算法  

Sketch Generation Algorithm Based on Gradient Correction and Multi Direction Texture Extraction

作　　者：梁义涛[1] 李亚飞 赵奎斌 李岚[1] 胡江汇[1] 

机构地区：[1]河南工业大学信息科学与工程学院,郑州450001

出　　处：《科学技术与工程》2017年第20期55-59,共5页Science Technology and Engineering

基　　金：国家自然科学基金(31171775);河南省教育厅高等学校重点科研项目(17A510008)资助

摘　　要：针对目前常见素描画自动生成算法中纹理方向单一、整体画面层次感不佳的问题,提出了一种基于梯度修正和多向纹理提取的素描画生成算法。该算法在图像分割基础上,针对不同分割区域提取亮度分布信息和梯度修正后的纹理方向分布信息,并利用这两种信息结合运动模糊滤波生成纹理图。最后将由霓虹处理生成的原图像的轮廓图与该纹理图合成最终的素描图。结果显示,本算法以运动模糊滤波代替线积分卷积来生成纹理,算法相对简化;以多向纹理和区域亮度信息更好地描述了整体的层次感、明暗对比度,视觉感官上获得了较好的素描效果。The common algorithms for automatic generation of sketch cause such problems as texture direction is poor and the overall screen layering is poor at present, a sketch generation algorithm base on gradient correction and multidirectional texture extraction was proposed.It is necessary to segment image in terms of the algorithm.On different segmentation regions basis, the algorithm extracts the luminance distribution information and the texture orientation distribution information after gradient correction, then the motion blur algorithm is applied to get texture map.Finally, original figure apply neon processing to generate contour map, the final sketch is obtained by fusing contour map and texture map.The algorithm takes the processing of motion blur to obtain texture map, instead of line integral convolution.The results show that the algorithm is relatively simplified.multidirectional texture and regional brightness information can describe the overall sense of hierarchy and contrast better, it is better sketch effect on the visual senses

关 键 词：图像处理 非真实感绘制 图像分割 多向纹理提取 梯度修正 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]