高效率視訊編碼(High Efficiency Video Coding)是以H.264/AVC為基礎的最新制定之影像壓縮標準。HEVC與H.264/AVC皆使用網路提取層(Network Abstraction Layer)使傳輸資料更有效率，但是若在網路傳輸過程中發生錯誤而造成封包(Packet)的遺失，則會因為失去資訊使影像品質降低，因此本論文主要探討當HEVC傳輸發生錯誤時，如何針對畫面內錯誤隱藏法對遺失的影像進行修補。本論文中比較了多種畫面內錯誤隱藏法，包含簡易的線性像素內插法與高計算複雜度的渲染技術等，經研究與探討後發現渲染技術中總變異量演算法與鄰近像素線性內插法相比其客觀效能僅高出約0.1dB，而在主觀角度中此兩種演算法對於修補不同的影像區塊各有優勢，因此本論文提出混合式演算法，藉由偵測影像之邊界與周圍預測模式，為欲修補之區塊選出較合適的錯誤隱藏法，實驗證實混合式演算法之平均效能較鄰近像素線性內插法高出0.5dB。;High Efficiency Video Coding (HEVC) is the latest standard for video compression based on the Advance Video Coding (H.264/AVC). Both HEVC and H.264/AVC provide a Network Abstraction Layer (NAL) to make transmission more efficient. However, if a transmission error occurs, this will cause packet loss, and the loss of information will cause the qualities degeneration. So this thesis focuses on using the spatial error concealment to recover the errors. In this thesis, we compare several spatial error concealment, including linear interpolation and inpainting technique with high computation complexity. Objectively, Total Variation algorithm can only achieve 0.5dB improvement on average compared to linear interpolation using nearest pixels. Subjectively, each algorithm has its own advantage in concealing the error. As a result, we propose a hybrid algorithm to choose the most suitable technique for the damaged block by detecting the edge and prediction mode of the block. The experimental results show that the hybrid algorithm can achieve 0.5dB PSNR improvement on average compared to linear interpolation using nearest pixels.
