摘要
针对现有视觉注意计算模型不适于处理维度大于4的多维图像的不足,将选择性视觉注意机制引入到多光谱遥感图像目标检测中,提出一种基于双四元数的视觉注意计算模型.将多维数据构建成双四元数的形式,利用其傅里叶变换的相位谱来提取显著性区域以用于显著目标检测;实现了多维数据的整体处理,并充分利用了频率域和空间域的信息.与传统的多光谱图像目标检测方法相比,该模型计算复杂度低、对各种参数设置的依赖性小.模拟数据与真实多光谱遥感数据的实验结果表明,文中方法具有良好的检测效果,同时对噪声具有较强的鲁棒性.
Since the existing computational models of visual attention are not suitable to process data with dimensions higher than four, we introduce selective visual attention mechanism into target detection on multispectral imagery, and propose a visual attention computational approach based on biquaternion. By transforming multi-dimensional data into biquaternion and exploiting the phase spectrum of biquaternion Fourier transform （PBFT）, saliency map is generated for salient target detection. The original multi-dimensional data can be incorporated as a whole, and features both in spatial and frequency domain can be extracted effectively. Compared with traditional multispectral target detection methods, our method has very low computational complexity and is not sensitive to parameter setting. Experimental results on simulated and real multispectral remote sensing data show that the proposed method has excellent performance in ship detection and is robust against white noise.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2011年第3期419-425,共7页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(61071134)
国家"八六三"高技术研究发展计划(2009AA12Z115)
关键词
视觉注意
双四元数
傅里叶变换
舰船检测
多光谱图像
visual attention
biquaternion
Fourier transform
ship detection
multispectral imagery
