结合区域检测特征点与极谐变换的图像拼接

IMAGE STITCHING COMBINING REGION DETECTION FEATURE POINTS AND POLAR HARMONIC TRANSFORM

提出一种结合区域检测特征点与极谐变换的图像拼接算法。首先,利用区域检测得到图像中的显著性区域,利用哈里斯(Harris)角检测器提取图像中的特征点,然后计算得出特征点圆形邻域的PHT特征矢量,并通过相似性变换原则消除误匹配对。最后利用正确的匹配对得出映射模型,并利用加权平均法进行图像融合。实验证明该算法提高了图像的拼接效率和准确性。

We propose an image stitching algorithm which combines the region detection feature points and the polar harmonic transform. First,the significant region in the image is detected by region detection,and the feature points are extracted from the image using Harris angle detector. Then the PHT feature vectors are computed over the circular neighbouring area around the feature points,and the mismatching pairs are eliminated according to the similarity transformation principle. After that,the mapping model is derived using the correct matching pair, and at last the weighted average method is used to make image fusion. Experiments prove that the algorithm improves the efficiency and accuracy of the image stitching.