標題: | 利用多重大小的自動回歸模型在多重描述編碼 Using Auto-Regressive Model with Multiple Training Window Sizes in Multiple Description Coding |
作者: | 王敬嚴 Wang, Ching-Yen 蔡文錦 Tsai, Wen-Jiin 資訊科學與工程研究所 |
關鍵字: | 錯誤隱藏、多重描述編碼、自動回歸模型;error concealment, multiple description coding, auto-regressive model |
公開日期: | 2012 |
摘要: | 由於視訊串流在近年來越來越受到使用者歡迎,錯誤修正的方法變得越來越重要。而多重描述編碼是一個有效的方法,常常用來克服不穩定的網路環境以及做錯誤補正。
在傳統的自動回歸模型裡,訓練窗格均採用固定的大小。在本篇論文中,我們提出一個利用多種不同大小訓練窗格的自回歸模型。而且利用各張最佳的視窗範圍大小的資訊,來加強多重描述編碼錯誤補償的能力。在我們的多重描述編碼結構中,我們擁有兩個描述子,一個有全部的奇數幀,另一個則有全部的偶數幀。兩個均使用H.264標準壓縮。在解碼的時候,我們可以利用自回歸模型以及對應的訓練窗格大小的資訊來進行遺失幀的補償。
從實驗結果顯示,我們的提出的方法與其他的方法相比,在主觀以及客觀的品質上都有較好的表現。 Since network video streaming have become popular in recent years, error resilient technique is more important. Multiple description video coding is one of well-known error resilient methods to cope with the network erroneous transmission in various networks environments. In conventional auto-regressive model, fixed training window size is adopted. In this thesis, we design a multiple description coding which adopts an auto-regressive model with multiple training window sizes to enhance the error resilience. In our MDC structure, we encode a video stream into two descriptions; one description contains all odd frames and the other contains all even frames. Both are encoded according to H.264/AVC standard. In the decoder side, we recover missing frames by using auto-regressive model with selected training window sizes. According to the experimental results, the proposed method outperforms other methods in both objective and subjective quality. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#GT070056086 http://hdl.handle.net/11536/71999 |
Appears in Collections: | Thesis |