動態影像區域分割之研究

Abstract

隨著影像科技的進步與人類生活需求，人們對動態影像的依賴性相對的提高，而對動態影像的應用範圍也明顯的擴大。例如動態影像中物體的分割、動態影像中物體的合成及動態影像的壓縮等應用。所以，為了滿足動態影像的相關應用，我們必須先發展一套理想的動態視訊區域分割技術，作為相關應用的前置處理。對於動態影像區域分割的技術而言，最重要的是要對區域或物體做一個明確的定義，適當的定義將對區域分割的研究會有很大的助益。 在本論文中，提出一個動態影像區域分割的方法。其中最主要是對每張頁框做靜態影像特性上的分析，依照區塊屬性，對頁框中的區塊和像素做初步分類，由上特性再對影像頁框做區域成長、區域分裂和區域合併，以達到區域分割的效果。對於頁框和頁框之間，區域的追蹤，我們去做運動向量的估算，再對區域做修正。在論文的最後，我們做了電腦模擬實驗，區域追蹤的結果並非如預期的完美。未來，區域追蹤的方法仍舊有很大的發展空間可以繼續努力加以改善。

With the advance of technology and the increasing demand in daily life, modern people have greatly depended on motion video, which is extensively employed in the segmentation, synthesis, and compression of objects. Therefore, to employ the motion video to the full, we have to develop an ideal set of segmental technology as a pre-process. For the segmental technology, the most important is to precisely define the region or the object. A clear definition will be helpful to the research in the regional segmentation. In this thesis, we focus our effort on segmentation. Firstly, we find some characteristics of pixels or blocks such that we can classify them. Then, the key techniques we use for region segmentation contain three stages including region-growing, region-splitting and region-merging. Finally, we estimate the motion vectors of regions and do some shape fixing for the estimated regions of next constructed frame under the region tracking stage. But the tracking results do not perfectly approach what we anticipate. In the future, it is of interest to find some method to improve the performance.