为实现两幅或多幅图像的拼接需要提取两幅图像中的特征点，并对所提取的特征点进行精确匹配。本文在应用SIFT方法提取待匹配图像中的特征点并对提取的特征点进行粗匹配基础上，对基于距离约束的特征点的精确匹配方法进行了研究。消除误匹配点对共分两个步骤：1）在一幅图像中取一个特征点计算与该特征点欧式距离最小的两个特征点，并取欧式距离最小值；利用粗匹配点，在待拼接的另一幅图像中同样求相应点间欧式距离并取最小值；若两值中的较小值与较大值之比大于设定阈值，则初步得到精确匹配点对。2）在一幅图像中顺序取初步判定位误匹配的特征点，计算该特征点与精确匹配特征点间欧氏距离，并取距离最小的两个特征点；在待拼接的另一幅图像中同样求相应匹配点间欧式距离并取最小值，两值中的较小值与较大值之比与设定的阈值比较可得到精确匹配点对。实验证明了该方法的有效性。%Image mosaic is a technique used to make up a high resolution image with two or more images. In the process of image mosaic the features should be acquired and matched accurately. The SIFT algorithm is applied to acquire the feature points of the image. The paper is focused on the accurate matching of the acquired features. Two steps are need in eliminating the miss-missing feature points. 1) Getting a feature point in one image, and finding the nearest two feature points in the same image. A minor length can be obtained. In the same way, the distance can be calculated applying the matched points in the other image, and another minor length can also be gotten. When the minimum length value is divided by the other value, the answer is acquired. If the answer larger than the threshold, the matched feature points are considered as the accurate matched point. Thus, only parts of accurate matched points can be ensured. 2) For acquiring more accurate matched points, a further procedure should be performed. The calculating procedure is similar to the first one. However, the gotten feature point is the mismatched point, and the nearest two feature points are the accurate matched points. Furthermore, the value of the threshold is different from the first step. The experiment results show that the proposed method can eliminate the mismatch feature points effectively.
展开▼
