現今通訊科技的發達,造成各類視訊科技的蓬勃發展,為了有效地傳輸、儲存與檢索視訊資料,發展一套理想的動態物件分割技術顯得特別重要,無論如何影像物件分割技術,至今仍是一個重要且尚待解決的問題。在本論文中,我們提出一個動態影像物件分割的方法,為了分割出有意義的物件,此論文主要是根據時間及空間的相似作為分割的依據,不同於傳統的分割方法只考慮空間的相似(包括位置、與色彩),此論文另提出物件切割還有所謂的時間相似度,即物件內的區塊在不同的時間內,所呈現的運動狀態也必須相似,因此所切割來的結果比傳統方法好了許多,其整體架構可以分成三大部分,包括a.過度分割(Over-Segmentation)、b.空間相似度的比較、c.時間相似度比較與合併,其中第一部份對一張影像分割成若干區塊,第二部分對每兩相鄰區塊作相似度的比較,第三部份則是對符合第二部分相似性條件的相鄰的區域作時間相似度合併,藉以達成動態影像物件的分割。實驗結果證實我們所提出方法的可行性,討論此方法的限制性以及未來的相關應用。
This paper proposes a technique for spatio-temporal segmentation to identify the objects present in the scene represented in a video sequence. This technique processes two consecutive frames at a time. A region-merging approach is used to identify the objects in the scene. No a priori knowledge is assumed about the number or characteristics of objects in the scene. First, Starting from an over-segmentation of the current frame. Second, a spatio-temporal similarity estimation. Finally, the objects are formed by iteratively merging regions together. Regions are merged based on their mutual spatio-temporal similarity. The spatio-temporal similarity measure takes both temporal and spatial information into account, the emphasis being on the former. We propose a novel measure to assess the spatio-temporal similarity between regions. This measure combines temporal and spatial information. The information about spatio-temporal similarity between regions is represented in the form of an adjacent region relation table. The region-merging process is based on a weight, adjacent region relation table. Experimental results on different types of scenes demonstrate the ability of the proposed technique to automatically partition the scene into its constituent objects.