基于特征点的图像配准子系统的研究与实现

摘要

本文针对图像配准,提出了一种能够应用硬件加速器实现的图像匹配处理算法,同时推出了相应的硬件架构.本算法基于图像的特征点的提取与特征点邻域的频谱匹配.同时,本算法可以用硬件逻辑单元实现,算法实现系统不需要加入数字信号处理器(Digital Signal Processor,DSP)或除法器.由软件仿真结果证明,本算法在匹配一定图像变换范围内的多幅图像时效果较好.在硬件设计中,本子系统综合考虑了系统的速度与硬件的开销,将二者控制在了可接受的范围.由FPGA逻辑综合结果可知,本子系统可以实现高速的图像处理,能够应用在实时监控的领域.%This paper focuses on the need for image matching. An algorithm that can be implemented into the hardware accelerator is proposed in this passage, and the hardware frame is given as well. The algorithm is based on the interest points and frequency spectrum comparison. It can be implemented into hardware without digital signal processors (DSP) or dividers. Proved by the simulation results, the performance of this algorithm is quite good in a certain range of image transformation. The hardware consumption is controlled in a reasonable amount by being designed in a speed-consumption compromising approach. According to the results of synthesis, the subsystem can meet the requirement of high speed processing, which allows the system to be used in many real-time imaging areas.