Automatic montage of retinal images in adaptive optics confocal scanning laser ophthalmoscope

摘要

We introduce an automatic technique to stitch retinal images recorded with an adaptive optics confocal scanning laser ophthalmoscope (AOSLO). The AOSLO images whose field size is 1.5 deg are recorded while the fixation target is moving according to a square pattern so as to cover a wide field of view. Principal components analysis–scale invariant feature transform is used to extract and match point features of the retinal images. With the matched point features, the affine transformation is used to stitch the images to generate eight montaged images. Then, the eight images are stitched together to construct a wide-field image by a cross-correlation algorithm. An image blending algorithm is also developed to reduce seam artifacts. With the proposed algorithms, a montage whose field size is approximately 10.5×9.6 deg is constructed.