由於近年來多媒體技術的蓬勃發展,許多應用例如多對一系統需要低耗電、簡單便宜的編碼裝置,分散式訊源編碼 (Distributed Source Coding,簡稱 DSC) 即是用來解決這類問題之最佳方案。如果將此應用於視訊處理上,即稱為分散式視訊編碼 (Distributed Video Coding,簡稱 DVC) 系統。一般 DVC 系統中會將影像分為 Intra、WZ (Wyner-Ziv) 兩種模式:Intra Mode為傳統 H.264 編碼方式;而 WZ Mode 會根據 Intra Mode 的資訊,將編碼端的複雜度轉移到解碼端,以達成上述多對一系統的條件。兩種模式的選擇會依據不同的影像而有不同的判斷,其判斷的結果將會對其傳輸碼率與影像品質有很大的影響,而傳統的 DVC 是利用 SAD (Sum of Absolute Differences) 值與 VAR (Variance) 並設下門檻值作為模式判斷的準則。 本篇論文目的在於改進傳統的區塊模式決策器 (Block Mode Decision,簡稱 BMD),提出 Lagrange Cost Function 作為 BMD 判斷的依據,與傳統 DVC 利用門檻值不同之處,在於 Lagrange Cost Function 使用位元傳輸率 (Rate) 與失真 (Distortion) 作為判斷的參數,對減少傳輸位元傳輸率與提高影像品質這兩項目標,做出最佳的選擇。而主要的研究,在於分析影響 Rate 與 Distortion 之參數,並建立 Lagrange Cost Function 之 Model,完成區塊模式決策器。模擬結果顯示,在 Foreman 與靜態影像 Hall Monitor 上,能夠比傳統區塊模式決策器有更好的效果。
The video technology is developing quickly in recent years, the multi-to-one applications which need low power consumption, low complexity and cheap encoder are noticed. The theory "Distributed Source Coding (DSC)" is used to solve the encoder problem, and it is called "Distributed Video Coding (DVC)" in the video applications. The video is divided into Intra and Wyner-Ziv (WZ) modes in DVC system, Intra is the coding of H.264, and WZ will translate the complexity from the encoder to the decoder according to the information of Intra mode for the multi-to-one applications. The selection of two modes have a great influence to the bit-rate and quality, and the traditional mode decision uses threshold value of Sun of Absolute Differences (SAD) and Variance (VAR) as the criterion. This thesis proposes a new Block Mode Decision (BMD) for DVC system based on Lagrange Cost Function, it have the difference to traditional BMD that uses the rate and distortion as the criterion, focuses on the two goals which is to reduce the rate and increase the video quality, and Lagrange Cost Function will make a best choice between rate and distortion. The main research is to find and analyze the parameters of rate and distortion, and build the model of Lagrange Cost Function to accomplish the new BMD. Simulation results show that Cost Function have better performance than the traditional BMD in sequence Foreman and low motion sequence Hall Monitor.