基于编辑传播的多源图像表观迁移

Multi-source Image Appearance Transfer Based on Edit Propagation

已有的图像表观迁移方法采用基于块的匹配技术,要求源图像和目标图像有较高的视觉相似性.然而,很多情况下用户很难轻易地找到符合要求的输入图像,从而大大限制了其应用.另一方面,已有的方法大多采用单一的迁移源,无法较好地处理复杂图像的表观迁移.针对以上问题,本文提出了一种基于编辑传播的多源图像表观迁移方法.首先,采用基于编辑传播的笔画交互系统在源图像和目标图像中进行简单的标记.提出一种新的笔画自动修正方法对用户标记的笔画进行分析处理,获得精确的笔画标记.然后,将修正后的笔画标记作为编辑传播中的控制样本,找到图像中所有与笔画标记相似的区域.为了解决复杂图像表观迁移问题,本文设计了基于多源图像的自动迁移算法.最后,结合颜色迁移和表观迁移技术,将对应的颜色或纹理信息迁移到目标图像中,得到理想的表观迁移结果.试验结果验证了本文方法的有效性.

Traditional appearance transfer methods automatically took advantage of dense patch matching method and required the source images and exemplar images had higher visual similarity. However, such pairs were rare. Furthormore, existing methods used a single source image, which were unable to deal with complex image appearance transfer. To solve the above problems, a multi-source image appearance transfer was pro- posed by editing propagation. Firstly user intuitively built regions, which were likely to have similar color sta- tistics by strokes and general precise stokes by a new automatically stroke correction method. Then edit propa- gation method was employed to find all the pixels with the similar appearance. To deal with the complex im- age, a multi-source image based automatically transfer method was introduced. Finally, our method performed color transfer or texture transfer between designated regions and generated the result of appearance transfer. Various experimental results verified the effectiveness of the proposed method.