Systems and methods for spoof detection based on gradient distribution
View Patent ↗A system and method for performing spoof detection are disclosed. The method includes: receiving an input image of a biometric; generating a first filtered image by applying a first convolution to the input image based on a first convolution kernel; generating a second filtered image by applying a second convolution to the input image based on a second convolution kernel; computing a gradient residual image by subtracting, for each pixel location of the first and second filtered images, a pixel value in the second filtered image from a corresponding pixel value in the first filtered image; applying a density estimation procedure to the gradient residual image to identify areas of varied density; and, determining whether the input image is a replica of the biometric based on results of the density estimation procedure.
1. A method for spoof detection, comprising:
receiving an input image of a biometric;
generating a first filtered image by applying a first convolution to the input image based on a first convolution kernel;
generating a second filtered image by applying a second convolution to the input image based on a second convolution kernel;
computing a gradient residual image by subtracting, for each pixel location of the first and second filtered images, a pixel value in the second filtered image from a corresponding pixel value in the first filtered image;
applying a density estimation procedure to the gradient residual image to identify areas of varied density; and
determining whether the input image is a replica of the biometric based on results of the density estimation procedure,
wherein applying the density estimation procedure comprises:
for each pixel location of the gradient residual image, summing an amount of gradient present in a window around the pixel location to generate a density value of the window around the pixel location;
computing an average density value of the density values of the windows;
determining a count of a number of windows that have a density value that deviates from the average density value by a threshold amount; and
identifying areas of varied density based on the count.
2. The method of claim 1 , wherein the window around each pixel location is a square-shaped window that overlaps at least one other window corresponding to a different pixel location.
3. The method of claim 1 , wherein the first convolution comprises a Gaussian blur, and the first convolution kernel comprises a standard deviation of a Gaussian distribution.
4. The method of claim 1 , wherein determining whether the input image is a replica of the biometric is further based on applying a density estimation procedure to one or more enrollment images of the biometric.
5. The method of claim 1 , wherein the biometric comprises a fingerprint of a finger, and the replica comprises a gelatin mold, a graphite mold, or a wood glue mold of the fingerprint of the finger.
6. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, causes a computing device to perform spoof detection, by performing steps comprising:
receiving an input image of a biometric;
generating a first filtered image by applying a first convolution to the input image based on a first convolution kernel;
generating a second filtered image by applying a second convolution to the input image based on a second convolution kernel;
computing a gradient residual image by subtracting, for each pixel location of the first and second filtered images, a pixel value in the second filtered image from a corresponding pixel value in the first filtered image;
applying a density estimation procedure to the gradient residual image to identify areas of varied density; and
determining whether the input image is a replica of the biometric based on results of the density estimation procedure,
wherein applying the density estimation procedure comprises:
for each pixel location of the gradient residual image, summing an amount of gradient present in a window around the pixel location to generate a density value of the window around the pixel location;
determining a window having a smallest density value; and
identifying areas of varied density based on the density value of the window having the smallest density value.
7. The computer-readable storage medium of claim 6 , wherein the window around each pixel location is a square-shaped window that overlaps at least one other window corresponding to a different pixel location.
8. The computer-readable storage medium of claim 6 , wherein the first convolution comprises a Gaussian blur, and the first convolution kernel comprises a standard deviation of a Gaussian distribution.
9. The computer-readable storage medium of claim 6 , wherein determining whether the input image is a replica of the biometric is further based on applying a density estimation procedure to one or more enrollment images of the biometric.
10. The computer-readable storage medium of claim 6 , wherein the biometric comprises a fingerprint of a finger, and the replica comprises a gelatin mold, a graphite mold, or a wood glue mold of the fingerprint of the finger.
11. A device, comprising:
a biometric sensor; and
a memory storing instructions; and
a processor configured to execute the instructions to cause the device to:
receive, from the biometric sensor, an input image of a biometric;
generate a first filtered image by applying a first convolution to the input image based on a first convolution kernel;
generate a second filtered image by applying a second convolution to the input image based on a second convolution kernel;
compute a gradient residual image by subtracting, for each pixel location of the first and second filtered images, a pixel value in the second filtered image from a corresponding pixel value in the first filtered image;
apply a density estimation procedure to the gradient residual image to identify areas of varied density; and
determine whether the input image is a replica of the biometric based on results of the density estimation procedure,
wherein applying the density estimation procedure comprises:
for each pixel location of the gradient residual image, summing an amount of gradient present in a window around the pixel location to generate a density value of the window around the pixel location;
computing an average density value of the density values of the windows;
determining a count of a number of windows that have a density value that deviates from the average density value by a threshold amount; and
identifying areas of varied density based on the count.
12. The device of claim 11 , wherein the window around each pixel location is a square-shaped window that overlaps at least one other window corresponding to a different pixel location.
13. The device of claim 11 , wherein the first convolution comprises a Gaussian blur, and the first convolution kernel comprises a standard deviation of a Gaussian distribution.
14. The device of claim 11 , wherein determining whether the input image is a replica of the biometric is further based on applying a density estimation procedure to one or more enrollment images of the biometric.
15. The device of claim 11 , wherein the biometric comprises a fingerprint of a finger, and the replica comprises a gelatin mold, a graphite mold, or a wood glue mold of the fingerprint of the finger.