利用馬可夫隨機域進行輪廓追蹤之景物切割技術

Abstract

隨著數位攝影機的普及，多媒體編輯和分析逐漸引起專家以及一般消費者的興趣。而對於多媒體編輯和分析，景物切割是一項相當重要且基礎的工作。使用者提供給景物切割系統關於前景及背景的提示，然後系統將景物從影片裡分離出來。前人所提出的方法常需要電腦長時間運算或是相當大量的使用者互動，而有些方法為了降低問題的複雜度而假設影片是由景物及靜態背景所組成，但是這些方法的應用卻因此而受到了相當限制。因此，我們提出一個新景物切割系統，整合以馬可夫隨機域為基礎的輪廓追縱法以及圖切景物切割法以減少運算量以及人為提示的需求。本系統利用輪廓追縱傳遞物件的形狀，再利用圖切演算法達到精確的切割。實驗結果顯示我們的演算法能用更少的關鍵影格且更有效率的切割影片裡的物件，因此降低了對使用者的互動的需求。

Video object segmentation is a critical task in multimedia editing and analysis. Normally, the user provides some hints of foreground and background, then the target object is extracted from the video sequence. Most previous methods are computation-expensive or labor-intensive, and some approaches that assume static background have limited applications. In this thesis, we proposed a novel video segmentation system that integrates MRF-based contour tracking with graph-cut image segmentation to reduce the computational cost and the user interaction. The contour tracking propagates the shape of the target object, and the graph-cut refines the shape to improve the accuracy of video segmentation. Experimental results show that our algorithm can efficiently segment the video sequence with less key-frames and hence less user interaction.