H.263视频编解码器在TM1300上的实现与优化

The Implementation and Optimization of H.263 Video Codec Based on TM1300

介绍了ITU_TH .2 6 3编解码算法原理和TM 130 0定点DSP芯片。针对该芯片的硬件结构特点 ,设计了一套运行于TM 130 0之上的实时视频信号采集、视频编解码、视频输出系统的可行方案。讨论了H .2 6 3实时视频编解码器在TM 130 0上定点实现的关键技术和难点问题 ,详细论述了H .2 6 3编解码算法的代码优化技术。测试表明 ,本文的实现及优化技术是可行的。

The ITU_T recommendation H.263 specifies a coded representation that can be used for compressing the moving picture component of audio_visual services at low bit rates. The basic configuration of the video source coding algorithm is based on Recommendation H.261 and is a hybrid of inter_picture prediction to utilize temporal redundancy and transform coding of the remaining signal to reduce spatial redundancy. TM1300 is a media processor for high_performance multimedia applications that deals with high_quality video and audio. TriMedia's DSPCPU family delivers exceptional performance and high_level language programmability for multimedia applications through the use of its VLIW architecture. TriMedia's VLIW architecture combines innovations in compiler and software design with advances in logic design. pSOSystem is a modular, high_performance real_time operating system designed specifically for embedded microprocessors. It provides a complete multitasking environment based on open systems standards. The feasible video codec running on TM1300 is designed according to the characteristic of the hardware structure. It includes real_time video signal sampling, video codec, video playing and so on. The emphatically key technologies and problem of the real_time implementation of H.263 video codec on TM1300 are discussed. The optimization technical is given in detail to the H.263 encoding and decoding algorithms for this hardware and compile system.This paper also describes several optimization ways,such as using custom operation, loop unrolling, disposing variable and function, decision tree grafting, compile and algorithm optimization,and various optimization methods supported by the TriMedia compilation system as well as techniques for exploiting the fine_grain parallelism of the TriMedia architecture. The test results are listed in the end.