基于主动轮廓探索的多源色彩迁移.pdfVIP

第 40 卷 第 4 期 自 动 化 学 报 Vol. 40, No. 4 2014 年 4 月 ACTA AUTOMATICA SINICA April, 2014 基于主动轮廓探索的多源色彩迁移 潘 杰 1 王雪松 1 程玉虎 1 摘 要 传统的多源色彩迁移算法常常利用欧氏色彩距离来分割目标图像, 由于色彩序列的模糊性与不确定性, 使得这种分 割极易出现色彩扭曲现象. 针对这个问题, 提出一种基于主动轮廓探索的多源色彩迁移算法. 首先, 为将目标图像的主体与背 景分离开, 利用一种主动进化的方法生成虚拟轮廓线, 并采用能量函数评价机制迫使虚拟轮廓线逐渐逼近实际轮廓线. 其次, 合理利用源图像与目标图像在 RGB、Gray 和 LMS 等不同色彩空间的表示、分割、转换, 实现其在 lαβ 空间的多源色彩迁移. 最后, 将在 lαβ 空间迁移得到的目标图像逆向操作后恢复为 RGB 显示. 单源与多源色彩迁移的对比、灰度化色彩通道的选择 以及各色彩空间不同色彩通道间的干涉性对比等实验验证了所提算法的合理性与有效性. 关键词 主动轮廓探索, 图像分割, 色彩空间, 多源迁移, 局部迁移 引用格式 潘杰, 王雪松, 程玉虎. 基于主动轮廓探索的多源色彩迁移. 自动化学报, 2014, 40(4): 713−720 DOI 10.3724/SP.J.1004.2014.00713 Multi-source Color Transfer Based on Active Contours Exploration PAN Jie1 WANG Xue-Song1 CHENG Yu-Hu1 Abstract Traditional multi-source color transfer algorithms often use Euclidean color distance to split the target image. However, the vagueness and uncertainty of the color sequence easily make the image segmentation prone to color distortion. In order to solve this problem, a multi-source color transfer based on active contours exploration algorithm is proposed. Firstly, for the purpose of splitting the object and background of the target image, an active evolutionary method is used to generate a virtual contour and an energy function evaluation mechanism is used to force the virtual contour gradually to approach the real contour. Secondly, representation, segmentation and transformation of the source and target images in different color spaces, i.e., RGB, Gray and LMS, are exploited to transfer color style from source images to the target image in the lαβ color space. Finally, the processed image in lαβ color space is operated backwards and is displayed in the RGB space. Simulation studies involving comparison of single-source and multi-source color transfers, selection for optimal color channel in Gray space, and analysis of mutual interference between color channels in different color spaces have verified the rationality and effectiveness of the proposed algorithm. Key words A