作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2010, Vol. 36 ›› Issue (19): 219-221. doi: 10.3969/j.issn.1000-3428.2010.19.077

• 图形图像处理 • 上一篇    下一篇

基于学习的烙画图像特征重构

王 东1,2,周世生1   

  1. (1. 西安理工大学印刷包装工程学院,西安 710048;2. 浙江工业大学之江学院,杭州 310024)
  • 出版日期:2010-10-05 发布日期:2010-09-27
  • 作者简介:王 东(1968-),男,讲师、博士研究生,主研方向:数字图像处理,色彩管理;周世生,教授、博士、博士生导师
  • 基金资助:
    陕西省2009年“13115”科技创新工程基金资助项目(2009ZDGC-06)

Learning-based Feature Reconstruction for Pyrography Painting

WANG Dong1,2, ZHOU Shi-sheng1   

  1. (1. Faculty of Printing and Packaging Engineering, Xi’an University of Technology, Xi’an 710048, China; 2. Zhijiang College, Zhejiang University of Technology, Hangzhou 310024, China)
  • Online:2010-10-05 Published:2010-09-27

摘要: 提出一种基于学习的烙画图像计算机仿真算法。在Hertzmann图像类比算法的基础上,使用融合轮选择算子的粒子群优化算法对处理过程进行加速,整个粒子群的当前最优位置采用轮选择的方法,能有效降低粒子群优化算法过早收敛于局部最优的机率,获得较传统近似最近邻域搜索更快的处理速度。为弥补Hertzmann算法中结果图像视觉效果上的纹理缺陷,在图像融合过程中,先将2幅输入图像转换到lαβ颜色空间,再进行点对点的加计算,进而实现图像视觉效果的增强。烙画图像的类比实验结果表明,该算法所获得的烙画仿真结果图像接近真实烙画图像效果。

关键词: 烙画, 学习, 轮选择, 粒子群优化

Abstract: This paper presents a novel learning-based approach for pyrography painting style simulation. Based on Hertzmann’s image analogies algorithm, it proposes a novel Particle Swarm Optimization(PSO) combined with the roulette selection operator to speed up the image processing. This algorithm selects the best position found by the swarm so far with roulette wheel selection method and the probability of premature converge to local minima is decreased. Compared with the Approximate Nearest Neighbor(ANN) search, it can obtain faster processing speed. In order to compensate for visual defects on the output texture in Hertzmann’s algorithm, it converts RGB signals to perception-based color space lαβ before the additional computing. Experimental results demonstrate the algorithm is efficient for pyrography painting simulation.

Key words: pyrography, learning, wheel selection, Particle Swarm Optimization(PSO)

中图分类号: