第21卷第11期 2013年11月
文章编号1004-924X(2013)11-2914-08
光学精密工程
Opties and Precision Engineering
Vol. 21No, 11
Nov.2013
无人飞行器超低延时高速并行图像压缩系统
李其虎1.2*，文运丰，贾瑞才！，肖嵩2
（1.中国电子科技集团公司第五十四研究所，河北石家庄050081：
2.西安电子科技大学综合业务网理论及关键技术国家重点实验室·陕西西安710071）
摘要：针对无人飞行器测控系统对图像压缩技术在超低延时与高速编码方面的特殊需求，研究并设计了一种全硬件并行图像压缩系统，该系统采用基于小波分析理论联合零游程编码与指数哥伦布编码技术完成对信源数据的压缩。利用小波分析去除像素相关性，引人最优量化方式，采用零游程编码结合指数哥伦布编码去除像素概率究余。算法在现场可编程门阵列（FPGA）中设计了一种并行处理模式·使得图像采集，变换与各个小波子代系数编码近似同步进行。实验结果表明，该压缩系统在型号为XILINX-VIRTEX-5-XC5VSX50T器件中可以同时满足4路分辨率为2048pixelX2048pixelX 12bit高清图像压缩，单路数据吞吐量达280Mpixel/s;针对1024pixel×768piexl格式图像编解码传输，延时低于100ms。关键调：无人飞行器；图像压缩：信源编码：并行处理；小波分析；超低延时
中图分类号：TN919.8：V279
文献标识码：A
doi:10.3788/OPE.20132111.2914
Rapid and parallel image compression system with ultra-low delay for unmanned aerial vehicles
LI Qi-hul-2-,WEN Yun-feng'.JIA Rui-cai',XIAO Song
(1.The 54th Research Institute of China Electronics Technology Group Corporation
Shijiazhuang050081,China;
2.State Key Laboralory of Integrated Service Nerwork,Xidian University,Xian 71oo7l,China)
*Correspondingauthor，E-mail:gucaso7@126.com
Abstract: A hardware parallel image compression system was designed according to the special re quirements of unmanned aerial vehicles (UAVs) for image compression in low delay and high speed coding. A compression algorithm based on the wavelet analysis theory combined with zero-run-length coding index Columbus was designed for source coding techniques, In the algorithm, the wavelet anal-ysis was applied to remove the redundancy of pixels, and the optimal quantization was introduced to remove the redundancy of pixel probability by zero-run-length encoding and exponential Columbus. The proposed algorithm was realized in a Field Programming Gate Array(PFGA) by a designed mode of the parallel processing, which makes the image sample, transformation and the coding of the each wavelet offspring coefficient executed in synchronization. Experimental results show that this com pression system can satisfy four pieces of high resolution images with a resolution of 2 048 pixel X 2 048 pixelX12 bit, and the capacity of data throughput achieves 280 Mpixel/s. Moreover, image co
收稿日期：2013-04-18；修订日期：2013-06-05.
基金项目：中国博士后科学基金资助项目（2013M541202）