基于多核计算的实时数字图像稳定  

Real-time Electronic Digital Image Stabilization System Based on Multi-core Computation

作　　者：周渝斌[1] 时永刚[2] 

机构地区：[1]上海工程技术大学机械工程学院,上海201620 [2]北京理工大学信息科学与技术学院,北京100081

出　　处：《计算机科学》2010年第4期269-273,277,共6页Computer Science

基　　金：上海高校选拔培养优秀青年教师科研专项基金(GJD08017)资助

摘　　要：多核计算已成为计算领域新的发展趋势。针对电子数字稳像系统中待处理的数据量和计算量都非常大的特点,结合双核PC平台的计算特点,采用了一种联合多种加快算法运行速度的处理方案,以满足系统的实时性要求。该方法结合图像视频数据处理的高度并行化、重复性处理的特征,在程序编制上,采用多线程并行处理方法,利用单指令多数据流(Single Instruction Multiple Data,SIMD)、流水线SIMD技术,在双核CPU上完成并行多线程任务分配和共享内存管理。并在快速搜索基础上,采用自适应多局部运动矢量估计消除运动物体干扰,用Kalman滤波分离了两种不同的运动信号,最终完成有害抖动信号的运动补偿。通过这些措施,在普通商用的双核PC机上实现了稳像实验系统的高效实时处理。The multi-core computing has become the new trend of computer industry. Electronic Digital Image Stabilization （EDIS） system involves both large amount of data and intensive computation. The real-time implementation of EDIS system entails rapid data processe with high parallelism and repeated computation, which are suitable for data parallel processe of image sequences. The algorithm solution was proposed to deal with high speed image stabilization to meet the requirements of real-time systems. According to the data processe characteristics of EDIS, efficient C＋ ＋ with inline SIMD （Single Instruction Multiple Data） and multi-thread programming were written and compiled on dual-core PC for real parallel running threads to achieve real-time performance in conventional PC. After the motion vectors acquired from fast search strategy, adaptive multi-local motion vectors avoided the moving objects interference. And the Kalman filter removed high-frequency image jitter motion while smooth global movements were retained for scanning and vehicle moving. The test results show that it is possible to implement a high efficient and robust real-time stabilization system on conventional PC with Multi-core at last.

关 键 词：多核计算 图像稳定 运动估计 并行处理 

分 类 号：TP311.11[自动化与计算机技术—计算机软件与理论]