针对斑点噪声对合成孔径(SAR)图像匹配算法的影响,提出了一种基于各向异性尺度空间的SAR图像匹配算法。首先,采用加性算子分裂算法解方案来构建各向异性尺度空间,在滤除斑点噪声的同时更好地保留图像细节;然后,在非线性尺度空间中提取特征点,并采用改进的SURF描述子描述特征,弱化斑点噪声对匹配的影响;最后,采用变换参数约束策略筛选匹配点对,提高匹配正确率。该方法既保持了同名点的精度还增加了同名点的数量,通过对不同极化、时相、波段以及不同视角下多种地物的匹配实验,验证了该方法的优越性。

Aiming at the effect of speckle noise in SAR image matching algorithm, a SAR image matching algorithm based on anisotropic scale space is proposed. Firstly, we use the additive operator splitting solution scheme to build the anisotropic scale space which can better preserve image detail while filter speckle noise at the same time; Then, we extract feature points in the nonlinear scale space, and use the improved SURF descriptors to weaken the effect of speckle noise; Finally points sets are matching based on transform parameter constraints policy filter. This method can maintain the matching accuracy while increase the number of points. Extensive experimental results on different polarization, phase and band SAR images demonstrated that our method can improve the matching accuracy.

TN957