基于特征点匹配的图像拼接方法

在对现有图像拼接方法进行研究的基础上,提出一种基于特征点匹配的图像拼接方法。采用对图像质量与光照等因素具有较强鲁棒性以及丰富匹配信息的尺度不变特征变换匹配算法对特征点进行提取和匹配,使用随机抽样一致性去粗算法求得两幅图像间的H矩阵初值,并运用迭代精炼算法求出H矩阵的精确值,利用摄像头预先拍摄好的固定角度的图像来得到转换角度的H矩阵,从而将均具有角度差的两幅图像调整至垂直拍摄的平面,最后根据尺度不变特征变换匹配算法及投影算法将调整后的两幅图像拼接。实验表明,本方法简单易行,可降低具有角度差的两幅图像拼接后的边缘变形,从而改善拼接质量。

Panorama mosaicking on scale-invariant feature transform

By analyzing difference of the panorama mosaicking methods available,a new one based on the scale-invariant feature transform（SIFT） algorithm is proposed.SIFT is robust to the image quality and illumination factors then extracting and matching abundant match information with this arithmetic.The initial value of H matrix is counted by Random Sample Consensus（RANSAC） and further conversion to accurate value of H matrix which counted by iterative refinement algorithm.Use the fix angle pictures which camera filming in advanced to get the transform angle H matrix,therefore,make the two images which have angle difference adjust to vertical shooting plane.Finally,mosaicking adjusted two images according SIFT and projection algorithm.Simulation proved that new method is simple and reduce the distortion at the edge field after two images mosaicking then improve its quality.