Eyelid shape estimation using eye pose measurement
Systems and methods for eyelid shape estimation are disclosed. In one aspect, after receiving an eye image of an eye (e.g., from an image capture device), an eye pose of the eye in the eye image is determined. From the eye pose, an eyelid shape (of an upper eyelid or a lower eyelid) can be estimated using an eyelid shape mapping model. The eyelid shape mapping model relates the eye pose and the eyelid shape. In another aspect, the eyelid shape mapping model is learned (e.g., using a neural network).
1. A method for estimating an eyelid shape mapping model, the method comprising:
under control of a hardware processor:
collecting a set of eye images of an eye; and
using the set of eye images to determine an eyelid shape mapping model that specifies a relationship between an eye pose of the eye and an eyelid shape of the eye,
wherein the eyelid shape mapping model comprises a mapping matrix that associates an eye pose coordinate system with an eyelid shape coordinate system,
wherein the eyelid shape is described by a set of eyelid shape parameters that specify a functional form for an eyelid boundary at an eyelid that is an upper eyelid of the eye or a lower eyelid of the eye,
wherein applying the mapping matrix to eye pose parameters in the eye pose coordinate system provides a functional decomposition of each eye pose parameter of the set of eyelid shape parameters expressed as a respective function of the eye pose parameters,
and wherein the set of eyelid shape parameters include a horizontal shift, a vertical shift, and a curvature of the eyelid.
2. The method of claim 1 , further comprising using the functional form to extract iris pixels for identity determination or verification.
3. The method of claim 1 , wherein the functional form comprises a pixel mask.
4. The method of claim 1 , further comprising using the functional form to initialize an eyelid search algorithm.
5. The method of claim 1 , further comprising using the functional form to exclude random guesses as candidates for membership in a final curve of a refined eyelid boundary determination.
6. The method of claim 5 , wherein using the functional form to exclude random guesses as candidates for membership in a final curve of a refined eyelid boundary determination comprises using the functional form to exclude candidates in a Random Sample Consensus (RANSAC) algorithm or a random subset-based fitting algorithm.
7. The method of claim 1 , wherein the eyelid shape mapping model comprises a parabolic curve having three parameters.
8. The method of claim 1 , wherein the eyelid shape mapping model comprises three quadratic polynomials, wherein inputs of the three quadratic polynomials comprises two angular parameters of the eye pose.
9. The method of claim 8 , wherein the two angular parameters are expressed as five parameters.
10. The method of claim 1 , wherein the eyelid shape mapping model is learned by a neural network comprising one or more hidden layers.
11. The method of claim 10 , wherein the one or more hidden layers comprise an output logistic layer.
12. The method of claim 1 , wherein the set of eye images comprise eye images of a left eye and a right eye.
13. The method of claim 12 , wherein the eye is the right eye, and wherein the eyelid shape mapping model is determined using a horizontal mirror image of the eye images.
14. A method for estimating an eyelid shape from an eye pose, the method comprising:
under control of a hardware processor:
determining, as a determined eye pose, an eye pose of an eye in an eye image; and
estimating an eyelid shape of an eyelid of the eye based at least partly on the eye pose, including applying an eyelid shape mapping model to the determined eye pose,
wherein the eyelid shape mapping model comprises a mapping matrix that associates an eye pose coordinate system with an eyelid shape coordinate system,
wherein the eyelid shape mapping model specifies a relationship between the eye pose of the eye and the eyelid shape of the eye, and
wherein the eyelid shape is described by a set of eyelid shape parameters that specify a functional form for an eyelid boundary of the eyelid,
wherein applying the mapping matrix to eye pose parameters in the eye pose coordinate system provides a functional decomposition of each eyelid shape parameter of the set of eyelid shape parameters expressed as a respective function of the eye pose parameters,
and wherein the set of eyelid shape parameters include a horizontal shift, a vertical shift, and a curvature of the eyelid.
15. The method of claim 14 , wherein the eyelid shape mapping model comprises a parametric relationship between the eye pose and the eyelid shape.
16. The method of claim 15 , wherein the eye pose is represented by a polynomial.
17. The method of claim 14 , wherein the eyelid shape mapping model comprises a neural network.
18. The method of claim 14 , further comprising:
re-determining, as a re-determined eye pose, the eye pose based on the eyelid shape; and
verifying consistency of the determined eye pose and the re-determined eye pose.