利用對稱式樹狀預測架構在AVC中有效實現視訊正轉、倒轉、快轉及隨機擷取

Abstract

我們提出了一種新的時間軸預測架構，將一串影像對稱式的拆解為 數個階層。這種架構提供了所有的VCR 功能如:隨機存取，快速正/倒 轉。我們將這種階層式的架構命名為對稱式樹狀預測架構。我們利用 AVC 提供的影像參照處理機制來設定並壓縮出符合AVC 標準的對稱式 樹狀預測架構二位元資料串流。對稱式樹狀預測架構在解壓縮端提供 了極有效率的快速 正/倒 轉，在快轉時，不需播放的影片可以略過不 解壓縮。對稱式樹狀預測架構只須壓縮一次，就可以單一的二位元資 料串流在解壓縮端提供所有的VCR 功能。 我們提供了相當有彈性的參 數設定，使我們可以任意變更其樹狀架構，使其有較好的壓縮效能或 是提供較低複雜度的壓縮。最後的實驗結果顯示，對稱式樹狀預測架 構其壓縮效能和一般壓縮方式相當接近，雖花了稍高的複雜度，但在 解壓縮端可以提供極低複雜度的VCR 實現。

We present a novel technique, hierarchical temporal domain decimated prediction structure, to provide a video bitstream with full VCR functionality – random access, fast forward, and fast backward. This hierarchical temporal domain decimated prediction structure, which we named as symmetric tree prediction structure, utilized the flexible reference picture management scheme that provided in AVC standard to generate an AVC compliant bitstream. It uses a tree-like prediction structure to provide efficient fast forward/fast backward playback functionality. No redundant picture will be decoded during fast playback. Only one-pass encoding and a single bitstream can provide both the forward and backward playback functionality. It also provides flexible parameters to make the prediction structure can be tuned with better coding efficiency or lower complexity. We also propose a decoder that can fully support the VCR functionality with low complexity. The simulation results show the symmetric tree prediction structure can provide full VCR functionality with similar coding efficiency and slightly higher coding complexity.