To overcome the shortcoming of SURF( speeded up robust features) feature matching algorithm, a speed-up method based on blocking is proposed. According to the sizes of template images and searching images, it is proposed to adopt the single blocking method that only splits the searching images and the double blocking method that splits both the template images and the searching images respectively. The former includes the method of simple K blocking,simple K + 1 blocking, scale K + 1 blocking as well as size and scale self-adaptive blocking, the double blocking method includes single block matching and multiple blocks matching. For different blocking method, the theoretic average matching times are analyzed and experimented, the results shows that the blocking method can speed up feature matching remarkably while keeping correct matching.%针对高实时性要求中SURF(speeded up robust features)特征匹配算法速度偏慢的缺点,提出一种基于分块的加速方法,根据匹配中模板图像与搜索图像的大小对比关系,分别采用只对搜索图像分块的单分块方法与对模板图像与搜索图像都进行分块的双分块方法,其中单分块方法包括简单K分块、简单K+1分块、尺度K+1分块及模板尺寸与尺度自适应的分块方法;双分块方法包括单块匹配与多块匹配.针对不同的分块方法分析了其理论上的平均匹配时间,并通过实验进行了对比,结果证明分块方法能够在保证正确匹配的同时大幅提高匹配速度.
展开▼
