Depth-aware blur kernel estimation method for iris deblurring
A method of estimating a blur kernel distribution for visual iris recognition includes determining a first mathematical relationship between an in-focus position of a camera lens and a distance between the lens and an iris whose image is to be captured by the lens. The first relationship is used to estimate a second mathematical relationship between the in-focus position of the lens and a standard deviation defining a Gaussian blur kernel distribution. A position of an eye of a living being at a future point in time is predicted. A focus position of the camera lens is adjusted based upon the predicted position of the eye. The camera lens with the adjusted focus position is used to produce an image of the living being's eye at the point in time. An actual position of the living being's eye at the point in time is sensed. The first relationship is used to ascertain a desired focus position of the lens based upon the actual position of the living being's eye at the point in time. The second relationship is used to calculate a standard deviation defining a Gaussian blur kernel distribution. The calculating is based upon a difference between the adjusted focus position and the desired focus position of the lens.
1. A method of estimating a blur kernel distribution for visual iris recognition, said method comprising the steps of:
determining a first mathematical relationship between an in-focus position of a camera lens and a distance between the lens and an iris whose image is to be captured by the lens;
using the first mathematical relationship to estimate a second mathematical relationship between the in-focus position of the lens and a standard deviation defining a Gaussian blur kernel distribution;
predicting a position of an eye of a living being at a future point in time;
adjusting a focus position of the camera lens based upon the predicted position of the eye;
using the camera lens with the adjusted focus position to produce an image of the living being's eye at the point in time;
sensing an actual position of the living being's eye at the point in time;
using the first mathematical relationship to ascertain a desired focus position of the lens based upon the actual position of the living being's eye at the point in time;
using the second mathematical relationship to calculate a standard deviation defining a Gaussian blur kernel distribution, the calculating being based upon a difference between the adjusted focus position and the desired focus position of the lens; and digitally unblurring the produced image by using the blur kernel distribution defined by the calculated standard deviation.
2. The method of claim 1 wherein the first mathematical relationship is determined based on empirical data.
3. The method of claim 1 wherein the second mathematical relationship is a linear relationship.
4. The method of claim 1 wherein a fixed position is used as a proxy for the predicted position of the eye.
5. The method of claim 1 wherein the first mathematical relationship is used in adjusting the focus position of the camera.
6. The method of claim 1 wherein the sensing step is performed using a depth sensor.
7. The method of claim 1 comprising the further step of determining whether the captured image needs to be modified based on the estimated blur kernel, which is realized by identifying a defocus-blur level of the captured iris image based on the estimated blur kernel.