摘要
图像配准是红外图像处理技术的前提,广泛应用于成像制导、光电侦查及工业实时监控等领域。考虑红外图像的实时配准需求,利用对数极坐标图像的尺度、转不变特性,将配准过程从直角坐标空间引入到对数极坐标空间,提出了基于双投影策略的配准算法,将二维空间的配准参数估计转换成两个一维曲线的空间距离估计。此算法在保证尺度变化及旋转的匹配精度条件下,轴向投影统计大大降低匹配计算量。
Image registration is a key technology applied in infrared guidance, reconnaissance, real time monitor and so on. Considering the need of real time and image in log-polar coordinate having scale and rotate unvariance property,a fast IR image registration algorithm based on coordinate dual-axis projection is proposed. Registration parameters estimation of image is transformed in projection curve distance estimation in new log- polar space. The new algorithm greatly speeds up processing from Cartesian coordinates to log-polar coordi- nates and ensures the reliability of matches. Theoretical analysis and experiment show that this method has better performance in both recognition speed and accuracy.
出处
《航天电子对抗》
2009年第3期22-24,50,共4页
Aerospace Electronic Warfare
关键词
图像配准
对数极坐标
红外凝视系统
image registration
log-polar coordinate linfrared staring system