形態變形法與等位函數變形法之比較

Abstract

現在有許多方法可以達到影像變形的目的。但是，這些方法在影像變形過程中常常伴隨著不適當的現象 – “鬼影”。在我的論文中，我們將會討論兩種可避免鬼影的方法。這兩種方法分別為數學型態差補法 (Morphological interpolations)以及等位函數集合法 (Level Set Methods)。更詳盡的說，在數學型態學中，我們比較數種型態差補法，同時我們以 blending 來代表等位函數法。在比較過型態差補法和 blending 的結果後，我們嘗試利用其中一種型態差補的影像來強化blending 的影像。實際上，我們可以發現到強化過的影像明顯的比原本的blending 影像明亮許多且變形過程一樣流暢。

There are several ways to achieve image morphing. But it always goes along with some improper phenomenon – “ghost effect” in morphing process. To avoid this phenomenon, in this paper, we will study on morphological interpolations and morphing by level set methods. More detailed, in our experiments, we implement several different interpolating ways to achieve morphological interpolations; and we implement blending to represent level set methods. After comparing the implement results of several kinds of morphological interpolations and blending, we make an attempt to enhance the blending morphed images reusing distance-based interpolation. Actually, we can observe that the results of enhanced blending are more brighten than original blending morphed images.