基于ARM的MPEG4 IDCT研究

Research on ARM-based MPEG4 IDCT

由于视频解码的运算复杂度很高,因此很多研究将微处理器上视频解码的IDCT,YUV/RGB转换等高运算部分硬件化实现,从而提高了成本,也提高了升级改进的难度。主要研究了MPEG4解码核心的IDCT部分的实现,选取了达到理论计算量下限的Loeffler算法的改进算法,在程序上进行了预判零技术和基于平台的优化。经过实际测试,在180MHz的ARM926EJS CPU上完成了QVGA格式MPEG4的30fps速度实时播放。

Due to the high complexity of video decoding, most researchers hardware those compute - intense part like IDCT, YUV to RGB to real- time decode video bitstream on embedded system, resulting in higher cost and difficulty in upkeep and upgrade. This paper mainly focuses on implementing IDCT during MPEG4 decoding. It chooses the Loeffler modified algorithm, which reaches the theoretical lower bound of multiplication number, and optimize the code by zero- block pre- detecting method and platform- based modification. By testing on 180MHz ARM926 board, the code reaches the requirement of 30fps of QVGA format MPEG4 file.