Highly optimized intra prediction architecture for high resolution application

This paper proposes a new intra prediction architecture for high resolution applications. The standard intra prediction has a data dependency for the pipelined processing. To enable the pipelined intra prediction architecture, processing order changing methods and additional processing schedulers are proposed. However, previous methods are not considered for the parallel processing which is a key issue of high resolution applications, such as High Definition videos or Ultra High Definition videos. The proposed intra prediction architecture has a new scheduler and two difference calculation processing units for the high performance parallel processing in intra 44 luminance prediction and intra 8x8 luminance prediction. The proposed architecture reduces processing time by about 43.40% compared with the standard and by about 20.10% compared with previous architecture.