基于SURF和伪Zernike矩的图像拼接算法研究

The Research of Image Mosaic Algorithm Based on SURF and Pseudo-Zernike Moments

基于特征点匹配的图像拼接方法对图像旋转、噪声等比较敏感,提出了基于SURF和伪Zernike矩的图像拼接方法,并利用改进的SURF算子获取图像中的特征点,计算以特征点为中心邻域窗口的伪Zernike矩,获得各个特征点邻域伪Zernike矩的Bray-Curtis相似性度量的初始匹配点对,利用RANSAC算法剔除伪特征点对,之后对输入图像作几何变换进行配准,融合重叠区域,获得良好的拼接图像.实验表明,改进的SURF特征检测算法提取的特征点均匀、准确、迅速,而且图像配准算法对平移、旋转以及噪声均具有鲁棒性.

The feature points matching image mosaic algorithm was found to be sensitive to rotations and noise, and an automatic image mosaic method based on SURF and Pseudo-Zernike moments algorithm was pro- posed. The improved SURF feature points detector was used to get the corners, then the pseudo-Zernike mo- ments defined on the interest point neighborhood were computed. Through comparing the Bray-Curtis similarity measure of the pseudo-Zernike moments to extract the initial feature points pair, then the spurious feature points pair were rejected by RANSAC algorithm, then by means of using the geometric transform of input ima- ges for registration, the overlapping region of two images was fused and the image Mosaic was finished. Im- proved SURF feature detection algorithm extract feature points are uniform, accurate and fast. The proposed method of image registration is robust to rotation, translation and noise.