关键词: H.264编码, 绝对误差和, 整数DCT变换, 变换绝对差值总和, 单指令多数据扩展指令集2（SSE2）

Abstract: H.264 video encoding standard adopts lots of new techniques. It has a significant performance benefit comparing with older standards in compression performance. However, it also has a considerable increase in encoder complexity, which limits the application it can be used for. For the Streaming SIMD Extensions 2（SSE2） instruction set which has the parallel computing power to improve the computer’s real-time processing of multimedia data. In this paper, instruction-level optimization of H.264 encoder is proposed by exploiting SSE2 instructions. The key time-consuming modules such as computing the sum of absolute difference（SAD） in integer pixel motion estimation, integer transform, quantization and computing the Hadamard transform of difference matrix and computing the sum of absolute transformed difference（SATD） in sub-pixel motion estimation. The experimental results show that the speed of corresponding module increases after being optimized while the same picture quality is achieved compared with the original encoder in H.264. The post-optimized coding speed of the encoder can satisfy the real-time requirement.

Key words: H.264 encoding, Sum of Absolute Difference（SAD）, integer DCT, Sum of Absolute Transformed Difference（SATD）, Streaming SIMD Extensions 2（SSE2）