利用Beamlet变换算法提取遥感图像直线

Line Extraction in Remote Sensing Image with Beamlet Transform

遥感图像处理中,直线特征是一种极为重要的特征,能够提供重要的视觉感知线索。本文利用一种Beamlet变换算法来提取遥感图像中的直线特征,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 Likelihood Ratio Testing)来检测判断符合条件假设的Beamlets,重建目标直线。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,这种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。

The straight lines feature is a very important feature and vision clue in Remote Sensing image processing. An algorithm using beamlet transform is proposed in this paper. Having dyadic and recursive subdivision on image, having integral on beamlets in each sub-square, the straight lines are approximated under GLRT（Generalized Likelihood Ratio Testing） and reconstructed. This method can overcome some kinds of noises in images and can extract exactly straight lines 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.