基于尺度不变特征的眼底图像自动配准与拼接

摘要

An automatic fundus image registration and mosaic algorithm based on scale invariant feature transform (SIFT) feature is presented to overcome the characteristics of low contrast, non-uniform illumination and the geometric distortion between different fields of view of the fundus images. Fundus images are enhanced by homomorphism filtering, then the SIFT features of fundus images are extracted and described using vectors to determine the matching feature point pairs between two images. Outlier point pairs are rejected using MLESAC algorithm employed perspective model, the transformation matrix is then computed according to purified matching point pairs between images. And finally, image registration and image mosaic are implemented by correcting the distorted image with a spatial transform. The registration and mosaic results of multiple images obtained by fundus camera show that the algorithm is robust and stable with registration accuracy up to pixel level, and high-precision automatic fundus image mosaic can be achieved.%针对眼底图像对比度低、光照不均匀、不同视场的图像间存在几何畸变等特点,提出了一种基于尺度不变特征的眼底图像自动配准与拼接算法.该算法分别提取同态滤波增强后的待配准眼底图像的尺度不变特征点,并用向量进行描述,确定相邻两图像特征点的匹配关系,在MLESAC算法中使用透视变换模型去除误匹配点对,计算匹配点对之间的变换矩阵,进行图像空间变换,完成配准和拼接.对实际眼底照相机获取的多幅图像配准与拼接结果表明,该算法具有很好的鲁棒性和稳健性,配准精度达到像素级,可以实现眼底图像的高精度自动配准与拼接.