利用Beamlet变换算法提取遥感图像线性特征

Linear features extraction from remote sensing image with Beamlet transform

利用一种Beamlet变换算法来提取遥感图像中的线性特征,通过对遥感图像按二进、递归进行划分,利用灰度信息,积分计算每一小块图像中的Beamlets,结合梯度信息,通过广义似然比检验GLRT(generalized likeli-hood ratio testing)来检测判断符合条件假设的Beamlets,重建线性目标。此算法可以克服图像中的干扰及噪声,适应复杂、低信噪比遥感图像环境,准确地定位出直线。实验结果表明,该算法具有较好的性能,可以应用到实际的遥感图像处理中,具有实用价值。

An algorithm using Beamlet transform was proposed. Having dyadic and recursive subdivision on image, having integral on Beamlet in each sub-square, the straight lines were approximated under GLRT and reconstructed. This method could overcome some kinds of noises in images and could extract exactly linear features even in low SNR images. The experimental results show that this method has a better performance and can be suited to remote sensing images processing.