基于PCA-SIFT特征检测的眼底图像拼接

The Fundus Image Mosaicking Based on PCA-SIFT Feature Detection

确定稳定的特征点是图像配准和拼接的重要环节。针对传统基于特征的拼接方法需要原始图像具有明显的角点或边界信息的弱点,将PAC-SIFT(Principal Components Analysis-scale invariant keypoints)主成分不变特征检测算法引入到眼底图像拼接中。该算法在划分的特征子区域内用梯度构造特征量,并用主成分分析法(PCA)降低特征维数。在拼接时,采用淡入淡出算法完成眼底图像的拼接。实验结果表明:该算法能够有效实现眼底图像的精确配准和平滑。

It is important to determine the keypoints of stability to the image registration and mosaicking.Based on the keypoints of the traditional method of Stitching,the original image need to have clear information on the border angle or Corner feature.So the PAC-SIFT feature detection method（Principal Components Analysis-scale invariant keypoints） is introduced into image feature detecting and matching.This algorithm divisiorys the sub-region with the gradient structure features and uses principal component analys is（PCA） to reduce dimensions.Finally, the method fade by fade is used to the Image Stitching of fundus images .Experimental results indicate that the proposed approach can effectively realize the fundus image registration and smooth.