An improved practical efficient implementation of ICT used in H.264

A parallel way to implement the 4times4 integer cosine transform (ICT), adopted in the emerging H.264 video coding standard, is presented. Our method can accelerate its calculation on general-purpose processors, such as Intel Pentium processors, by making use of MMX technology. The underlying principle lies in the fact that some general purpose processors can execute several independent operations at the same time. The proposed method first analyzes independent operations involved in ICT, and then uses two steps to realize parallel computation: preprocessing, which takes charge of reorganizing source input in a dedicated way for forthcoming processing; parallel computation, which is responsible for ICT calculation with MMX instructions. Compared with its realization in H.264 reference software, our study shows this method can speed up the calculation of ICT by about 60%