最新一代的視訊壓縮標準HEVC(High Efficiency Video Coding)將畫面內預測之角度模式提升至33個方向，使得預測畫面的方向性能夠更加接近原始畫面，而得到更好的壓縮效率。然而HEVC角度預測擁有單一方向性的缺點，對於擁有複雜紋理的畫面還有改進的可能性。因此本篇論文提出將原本使用於半色調法(Halftoning)的錯誤擴散法(Error Diffusion)應用於畫面內預測(Intra Prediction)之中。我們首先利用對梯度的調整控制畫面的平滑度，接著再利用錯誤擴散法將調整後的誤差擴散出去。錯誤擴散法和渲染方法一樣與周圍像素點有很高的相依性，因此能夠同時將雙邊參考像素資訊延續至預測畫面當中，並且運算複雜度較渲染方法低非常多。因此錯誤擴散法能夠在不增加過多編碼時間的情況下，有效改善含有複雜紋理之區塊的預測畫面，進而增進壓縮的效能。實驗結果表明，錯誤擴散法與HEVC畫面內預測之結合型演算法能夠降低0.5%的BDBR(Bjontegaard Delta Bit Rate)。;High Efficiency Video Coding, the latest standard of video compression, has improved the efficiency of Intra Prediction by making finer prediction image due to the increasing of Intra Angular Prediction modes, which now have 33 different directions. However, the characteristic of single directionality has not changed. In order to improve the prediction of images with complex texture, we apply Error Diffusion, which was originally a technique of Halftoning, to Intra Prediction. We control the smoothness of the image by adjusting the gradient between pixels, followed by diffusing errors to neighboring pixels. Hence, Error Diffusion has as high correlation with neighboring pixels as Inpainting does. It can improve the quality of prediction images by extending the information of bilateral reference pixels to them with only a little increase of time consumption. It was showed that the hybrid algorithm combined by Error Diffusion and HEVC Intra Prediction achieves 0.5% BDBR(Bjontegaard Delta Bit Rate) saving compared with standard HEVC Intra Coding by the experimental results.
