Image correction for ophthalmic images

1. A method of generating a correction algorithm, comprising:

obtaining a model of an eye, the model of the eye including:

a front portion comprising optics to mimic a cornea and a lens of a human eye; and

a rear portion comprising a generally hemispherical-shaped body to mimic a retina of the human eye, the rear portion including physical reference lines on an inside surface of the generally hemispherical-shaped body;

capturing images of the model of the eye using at least one image capturing device aimed at the model of the eye in which the at least one image capturing device is to be used when imaging an actual eye;

identifying vertices of the physical reference lines according to a given projection technique for displaying generally hemispherical-shaped body in a two-dimensional image in the captured images;

obtaining an idealized placement of the vertices of the physical reference lines according to the given projection technique;

performing bilinear interpolation to determine a first surface for correcting any pixel in the two-dimensional image according to a first difference between the identified vertices in the captured images and the idealized placement in an x-axis, and a second surface for correcting any pixel in the two-dimensional image according to a second difference between the identified vertices in the captured images and the idealized placement in a y-axis; and

storing the first surface and the second surface as part of the correction algorithm such that any pixel of an image captured according to how the image capturing device is to be used when imaging the actual eye is corrected using the first surface to adjust a given pixel in the x-axis and the second surface to adjust the given pixel in the y-axis.

2. The method of claim 1 , further comprising:

capturing human-eye images of the actual eye using at least one of a second set of image capturing devices;

applying the correction algorithm to the human-eye images; and

stitching together the corrected human-eye images to achieve a wide field of view image of the actual eye, the wide field of view image depicting a retina of the actual eye according to the projection technique.

3. The method of claim 1 , further comprising:

generating third and fourth surfaces based on a second projection technique; and

storing the third and fourth surfaces as part of a second correction algorithm related to the second projection technique.

4. The method of claim 3 , further comprising:

capturing human-eye images of the actual eye using at least one of a second set of image capturing devices;

applying the second correction algorithm to the human-eye images; and

stitching together the human-eye images corrected using the second correction algorithm to achieve a wide field of view image of the actual eye, the wide field of view image depicting a retina of the actual eye according to the second projection technique.

5. The method of claim 1 , further comprising:

subdividing a corrected image of a human eye into smaller portions, the corrected image corrected using the correction algorithm;

classifying the smaller portions for quality; and

storing the classified smaller portions for later use in a machine-learning algorithm.

6. The method of claim 1 , further comprising superimposing a partially transparent image of the rear portion of the model of the eye on a corrected image of a human eye, the corrected image corrected using the correction algorithm.