Information re-use and edge detection in intra mode prediction

In intra mode prediction, obtaining the best solution among all supported modes requires extremely high computation. Most fast search algorithms use a reduced search set to avoid this problem. Yet, these methods greatly degrade picture quality and increase bitstream size if very limited computational resource is applicable. In this paper, we proposed a scheme, which allows us to take extensive advantage of correlation within a video sequence. Prediction errors and modes of neighboring blocks are employed in our mode prediction scheme. The proposed method can either work alone or with existing direction detection methods to significantly reduce computation without losing much performance.