基于sift特征描述符的多尺度图像配准方法

A Multi-Scaled Image Matching Method Based on Sift Character Descriptor

研究图像处理,提高图像的精确性,在基于特征的图像配准中,特征描述符用来对两幅图像的特征进行相似性度量。合适的特征描述符对于建立图像之间的配准映射关系和提高配准精度具有重要意义。为了适应图像的尺度变化,提高配准算法的精度,引入多尺度Harris角点检测算法,并对一种基于sift特征的描述符在向量构造和采样区域等方面进行改进,最后通过对比特征描述符的相似性建立特征点间的匹配关系,并进行仿真。结果证明,算法能够适应图像的尺度变化,增加了描述符在图像具有噪声、旋转时的鲁棒性,提高了图像配准的精度。

In the methods of image matching based on the characters,the character descriptor is used to measure the two pictures＇ similarity.A good character descriptor plays a significant role in establishing the relationship between two images and increasing the precision of image matching.In order to adapt the image＇s scale changes,increase the precision of matching method,this paper leads the multi-scale Harris points detecting theory into the algorithm,and improved a character descriptor based on sift character for its vector designing,sampling area aspects and so on,at last establish the relationship between the angular points through comparing the similarity of the descriptor.The experiments prove that,this algorithm is able to adapt the pictures＇ scale changes,and enhanced the robustness of the descriptor when the image have noise and rotation changes,increased the precision in image matching.