耦合分水岭与形态学的特殊纹理影像配准算法

A registration algorithm of images with special textures coupling a watershed with mathematical morphology

针对现有算法在合成孔径雷达(synthetic aperture Radar,SAR)影像与光学影像配准时存在的效率和精度较低的问题,提出一种耦合标记控制分水岭与数学形态学的特殊纹理影像逐步求精的自动配准算法。首先,利用改进的标记控制分水岭算法分别提取影像中的水体特征,并进行二值化和数学形态学处理,以准确地提取水体区域;其次,提取水体质心用于图像间的粗配准,提升后续算法搜索效率;最后,基于优化算法搜索得到相似性测度最大时的最优变换参数,以此对待配准SAR影像进行空间变换,完成SAR影像与光学影像的精配准。实验结果表明,该算法耦合了图像分割与配准,在减少计算量的同时确保配准精度,有效地解决了灰度和分辨率差异大的SAR影像与光学高分辨率影像自动配准的难题。

Existing registration algorithms suffer low efficiency and accuracy in the registration of synthetic aperture Radar(SAR)and optical images.This study proposed a stepwise refinement-based automatic registration algorithm of images with special textures by coupling marker-controlled watershed segmentation and mathematical morphology.Firstly,the improved marker-controlled watershed algorithm was used to extract the features of water bodies from images,and then binarization and mathematical morphology were applied to accurately extract the water regions.Secondly,the centroids of water bodies were extracted for rough registration between images to improve the search efficiency of the subsequent algorithm.Finally,using the optimization algorithm,the optimal transformation parameters when the similarity measure was maximized were obtained and were used to carry out the spatial transformation of SAR images for image registration.In this manner,the precise registration of SAR and optical images was completed.The experimental results show that the algorithm proposed in this study that couples image segmentation with registration reduced calculation amount while ensuring the registration accuracy.Meanwhile,this algorithm effectively solved the difficulty in the automatic registration of SAR and optical high-resolution images that have large differences in gray levels and spatial resolution.