提出一种图像全局特征描述方法——梯度径向夹角金字塔直方图。该特征不受图像灰度级非线性变换和图像旋转的影响,具有较强的抗噪声能力。基于该特征的匹配算法能够对存在旋转变换的异源图像进行匹配。用SAR图像和可见光图像对匹配算法进行测试,结果表明基于梯度径向夹角金字塔直方图匹配方法的匹配成功率优于传统方法。由于对基准图进行离线预处理,同时避免角度空间搜索,算法的实时计算时间比传统方法小3个数量级。 更多还原
【Abstract】 A novel image feature named gradient radius angle pyramid histogram(GRAPH) is proposed.The GRAPH feature is invariable to nonlinear transform of grey level and image rotation,and is robust to noise.Based on GRAPH feature,we get a robust and fast image matching algorithm that can match rotated and shifted multi-sensor images.SAR images and optical images are used to test this matching method.The results show that the new matching algorithm has higher success rate than traditional method.As the re... 更多还原