【摘要】 介绍一种基于SIFT特征的图像拼接方法。它首先使用SIFT算法提取特征点及其描述符,然后分别为两幅待拼接图的特征描述符建立K-d树,并通过K-d树查找和确定初始匹配点对。接下来介绍和使用新提出的一种利用投影变换前后的相对位置关系实现匹配对精炼的方法。接着在匹配点对中选择包围区域最大的4对点,用以计算变换矩阵,最终得到拼接图像。实验结果证明,该方法有效。更多还原

【Abstract】 Image stitching has been widely used in many fields. It is introduced in this paper that stitching image based on SIFT. At first, SIFT is used to find feature points and its descriptors. Next, k-d tree of descriptors is build for two stitching images respectively and used to find original matching point pairs. Then, a new method creatively is put forward and applies to refine matching point pairs, which makes use of position relation before and after transforming. Afterwards,surround area biggest four point pairs are chosen from matching pair points to calculate transformation matrix. At last, using weighted melting to stitch two image together. According to experiments,the way is effective.更多还原