動態的的外形輪廓之壓縮方法

Abstract

在本論文中主要是探討不同時間上動態的外形輪廓之間的壓縮方法（Inter-Mode Contour-based Shape Coding），如何充分的利用時間上和空間上的相關性。 因此我們分析動態的外形輪廓之間的壓縮方法的流程之三個步驟：在前一個時間上找到一個參考的物件外形、計算兩者之間的誤差和壓縮此誤差。並就各個步驟分析並定義出動態的外形輪廓之壓縮方法的困難點和可能的解決方法。 再根據物件輪廓的三種不同表示法：相連的點序列、一維的訊號型態的序列和多邊形近似的表示法，探討不同的輪廓表示法在動態的外形輪廓之壓縮問題的考量點和達成最佳解的程度。並且分析探討各相關方法的問題及改善的技巧，提出了Modified fragment-divided和Modified DPCM baseline-based兩方法，能比原本的方法降低計算複雜度和提高壓縮率。

In this thesis, we address the problem of the inter-mode contour-based shape coding. We analyze the procedures of inter-mode contour-based shape coding, which are finding reference, evaluating difference, and coding difference, and indicate issues and possible solution for each procedure. Representation of contour includes connected pixel sequence, one dimension signal sequence, and polygon-approximation. According to different contour representation, we analyze issues and solution for each contour representation. Furthermore we study and improve related methods of inter-mode contour-based shape coding, and propose modified fragment-divided and modified DPCM baseline-based methods that take less computation complexity and higher coding gain.