分散式視訊編碼 (DVC, Distributed Video Coding) 和傳統視訊架構的視訊編碼應用取向不同,它的特色是使用低運算複雜度的編碼器。我們提出一個創新的分散式視訊編解碼器架構,在極大碼率下獲得了比以往DVC更好的效能。我們在Intra-Frame的處理上採用JPEG 2000 處理Key Frame,並且在Wyner-Ziv Frame的處理上捨棄LDPCA (Low-Density Parity Check Accumulate Codes) 的使用,改用全部CAVLC (Context-based Adaptive Variable-Length Code) 的編解碼方式。在極大碼率的情況下,我們提出的架構,在高畫質需求的應用上獲得更好的效能。以解析度352 x 288的Highway視訊序列為例,碼率大於1.73 bbp (bit per pixel)之後的一定碼率下,我們獲得47.3 dB以上高於Hybrid DVC的PSNR (Peak Signal-to-Noise Ratio) 值。而且編碼時間和解碼時間皆有顯著減少,以執行解析度176 x 144的視訊序列Foreman為例,平均的編碼時間和解碼時間分別是Hybrid DVC的43.8%和8.2%。
Distributed Video Coding (DVC) was proposed to meet the need of low complexity encoder. The background knowledge of DVC is based on two theories, Slepain-Wolf Theory and Wyner-Ziv Theory. DISCOVER (Distributed Coding of Video service) and Hybrid DVC are two of the most popular DVC architectures in recent years. In this Thesis, we proposed a novel architecture of DVC. At extremely high bit rate, our DVC has a better video quality in high resolution and fast motion sequences than DISCOVER and Hybrid DVC. In the sequence of 352 x 288 Highway, when its bit rate is larger than 1.73 bpp (bit per pixel), the PSNR (Peak Signal to Noise Ratio) is getting higher than 47.3 dB, achieving a better quality than Hybrid DVC. Moreover, executing time of our work is less than DISCOVER and Hybrid DVC. As to the sequence of Foreman 176 x 144, its average encoding and decoding times are respectively 43.8% and 8.2%, compared to Hybrid DVC.