Share Email Print

Proceedings Paper

A MPEG-4 encoder based on TMS320C6416
Author(s): Gui-ju Li; Wei-ning Liu
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Engineering and products need to achieve real-time video encoding by DSP, but the high computational complexity and huge amount of data requires that system has high data throughput. In this paper, a real-time MPEG-4 video encoder is designed based on TMS320C6416 platform. The kernel is the DSP of TMS320C6416T and FPGA chip f as the organization and management of video data. In order to control the flow of input and output data. Encoded stream is output using the synchronous serial port. The system has the clock frequency of 1GHz and has up to 8000 MIPS speed processing capacity when running at full speed. Due to the low coding efficiency of MPEG-4 video encoder transferred directly to DSP platform, it is needed to improve the program structure, data structures and algorithms combined with TMS320C6416T characteristics. First: Design the image storage architecture by balancing the calculation spending, storage space cost and EDMA read time factors. Open up a more buffer in memory, each buffer cache 16 lines of video data to be encoded, reconstruction image and reference image including search range. By using the variable alignment mode of the DSP, modifying the definition of structure variables and change the look-up table which occupy larger space with a direct calculation array to save memory space. After the program structure optimization, the program code, all variables, buffering buffers and the interpolation image including the search range can be placed in memory. Then, as to the time-consuming process modules and some functions which are called many times, the corresponding modules are written in parallel assembly language of TMS320C6416T which can increase the running speed. Besides, the motion estimation algorithm is improved by using a cross-hexagon search algorithm, The search speed can be increased obviously. Finally, the execution time, signal-to-noise ratio and compression ratio of a real-time image acquisition sequence is given. The experimental results show that the designed encoder in this paper can accomplish real-time encoding of a 768× 576, 25 frames per second grayscale video. The code rate is 1.5M bits per second.

Paper Details

Date Published: 21 August 2013
PDF: 7 pages
Proc. SPIE 8908, International Symposium on Photoelectronic Detection and Imaging 2013: Imaging Sensors and Applications, 89081F (21 August 2013); doi: 10.1117/12.2033614
Show Author Affiliations
Gui-ju Li, Changchun Institute of Optics, Fine Mechanics and Physics (China)
Wei-ning Liu, Changchun Institute of Optics, Fine Mechanics and Physics (China)

Published in SPIE Proceedings Vol. 8908:
International Symposium on Photoelectronic Detection and Imaging 2013: Imaging Sensors and Applications
Jun Ohta; Nanjian Wu; Binqiao Li, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?