Method and apparatus for determining eye position information

1. A method of determining eye position information, the method comprising:

identifying an eye area in a facial image;

verifying a two-dimensional (2D) feature in the eye area; and

performing a determination operation including,

determining a three-dimensional (3D) target model based on the 2D feature; and

determining 3D position information of an eye corresponding to the eye area, based on the 3D target model,

wherein the verifying includes,

determining a supervised descent method (SDM) model based on a local binary pattern (LBP), and

verifying the 2D feature in the eye area based on the SDM model, and

wherein verifying the 2D feature in the eye area based on the SDM model includes,

acquiring a sample area by measuring an eye area in a sample image, and

performing a repetitive training on the SDM model based on a sample feature measured in the sample area.

2. The method of claim 1 , wherein the determination operation includes,

obtaining a parameter associated with the 2D feature; and

establishing the 3D target model based on the parameter.

3. The method of claim 1 , wherein the determination operation includes,

obtaining a matrix based on the 2D feature and the 3D target model; and

determining the 3D position information based on the matrix.

4. The method of claim 1 , wherein the verifying comprises:

determining whether an eye is in a stationary state by comparing a first frame and a second frame of the eye area, the second frame being subsequent to the first frame; and

verifying the 2D feature in the eye area in response to a determination that the eye is not in the stationary state.

5. The method of claim 4 , wherein the determining whether the eye is in the stationary state further comprises:

determining previously determined 3D position information to be current 3D position information in response to a determination that the eye is in the stationary state.

6. The method of claim 4 , wherein the determining whether the eye is in the stationary state comprises:

calculating a normalized correlation coefficient of a previous image frame of the eye area and a current image frame of the eye area; and

determining that the eye is in the stationary state when the normalized correlation coefficient exceeds a threshold.

7. The method of claim 1 , further comprising:

converting the 3D position information to a 3D coordinate system of a display; and

adjusting or rendering a 3D image of the display based on a result of the converting.

8. The method of claim 1 , wherein the performing of the repetitive training comprises:

extracting an approximate feature of the sample area in an initial repetitive training; and

extracting a precise feature of the sample area in a subsequent repetitive training,

wherein the approximate feature includes at least one of a histogram of oriented gradients (HOG) feature, a multi-block local binary pattern (MB-LBP) feature, a speeded up robust feature (SURF), and an oriented FAST and rotated BRIEF (ORB) feature, FAST being features of accelerated segment test and BRIEF being binary robust independent elementary features, and

the precise feature includes at least one of an LBP feature, a Gabor wavelet feature, a discrete cosine transform (DCT) feature, and the BRIEF.

9. The method of claim 8 , wherein the performing of the repetitive training comprises:

extracting a feature of the sample area and performing training on an SDM model acquired during a previous repetitive training process in different-sized spaces;

comparing the SDM model on which the training is performed in each different-sized spaces to a feature of the sample area measured in advance; and

selecting an SDM model from a plurality of SDM models based on a result of the comparing and applying the selected SDM model to a subsequent repetitive training process.

10. The method of claim 1 , wherein the identifying comprises:

determining a position of an eye;

generating a virtual eye frame based on the determined position; and

acquiring the eye area in a facial image corresponding to a current frame based on the virtual eye frame,

wherein the determined position is associated with position information of the 2D feature.

11. An apparatus for determining eye position information, the apparatus comprising:

memory storing computer-executable instructions; and

one or more processors configured to execute the computer-executable instructions such that the one or more processors are configured to,

identify an eye area in a facial image,

verify a two-dimensional (2D) feature in the eye area, and

determine a three-dimensional (3D) target model, based on the 2D feature and determine 3D position information of an eye corresponding to the eye area, based on the 3D target model,

wherein the verifying of the 2D feature in the eye area includes verifying the 2D feature using a supervised descent method (SDM) module based on a local binary pattern (LBP), and

wherein the one or more processors are configured to excute the computer-executable instructions such that the one or more processors are further configured to,

determine a sample area by measuring an eye area in a photographed sample image, and

perform repetitive training on an SDM model based on a sample feature measured in the sample area.

12. The apparatus of claim 11 , wherein the one or more processors are configured to execute the computer-executable instructions such that the one or more processors are further configured to,

obtain a parameter by determining the 2D feature and establish the 3D target model based on the parameter;

obtain a matrix based on the 3D target model and the 2D feature; and

determine 3D position information based on the 3D target model and the matrix.

13. The apparatus of claim 11 , wherein the one or more processors are configured to execute the computer-executable instructions such that the one or more processors are further configured to,

compare a first frame and a second frame of the eye area and determine whether an eye is in a stationary state, the second frame being subsequent to the first frame,

the verifying of the 2D feature in the eye area including verifying the 2D feature in the eye area in response to a determination that the eye is not in the stationary state.

14. The apparatus of claim 13 , wherein the one or more processors are configured to execute the computer-executable instructions such that the one or more processors are further configured to determine previously determined 3D eye position information to be current 3D eye position information in response to a determination that the eye is in the stationary state.

15. The apparatus of claim 13 , wherein the one or more processors are configured to execute the computer-executable instructions such that the one or more processors are further configured to,

calculate a normalized correlation coefficient of a previous image frame of the eye area and a current image frame of the eye area, and

determine that the eye is in the stationary state when the normalized correlation coefficient exceeds a threshold.

16. The apparatus of claim 11 , wherein the one or more processors are configured to execute the computer-executable instructions such that the one or more processors are further configured to,

convert the 3D position information determined by the 3D position information determiner to a 3D coordinate system of a display, and

adjust or render a 3D image of the display based on the 3D position information.