基于活动轮廓模型的自动图形变形

图形变形算法中最费时的是在源图形和目标图形中确定变形特征基元；而到目前为止，大部分图形变形算法都是通过人工的方式来实现的。如果两幅源图形间的特征基元的确定能自动实现，这样不但可以减轻用户的负担，而且可以极大地节省变形算法所需的时间，同时提高了算法的效率。基于此考虑，本文提出了一种基于梯度向量流活动轮廓模型的自动图形变形算法，该算法利用基于梯度向量流的活动轮廓模型来自动精确地指定源图形和目标图形的变形特征点及其对应关系。

Graphics Morphing Based on Active Contour Model

This paper presents new solutions to the feature specification and warp function problems in image morphing. To reduce the burden of feature specification, we first adopt GVF snakes, which help animator to easily and precisely capture the exact position of a feature. They also reduce the work of an animator in establishing the feature correspondence between two image sequences. We next propose the use of B-spline to achieve one-to-one warps among feature point pairs. The algorithm is simpler and faster than previous warp generation methods. The resulting warps provide visually pleasing image morphing.