Facial liveness detection

1. A system comprising:

a processor;

an image analysis module coupled to the processor, wherein the image analysis module is to:

processing captured image to detect facial portion of a user within the captured image;

deriving a facial image from the processed captured image;

for the derived facial image, determining a set of specular features and texture-based feature vector, wherein for determining specular features, the image analysis module is to:

convert the facial image to a YUV color space to obtain a converted facial image; and

process the Y-channel of converted facial image to obtain the specular features, wherein the processing of the Y-channel further comprises:

normalizing the Y-channel;

performing a histogram normalization of the normalized Y-channel;

further performing an intensity transformation on the normalized Y-channel to obtain a transformed Y-channel; and

subtracting the transformed Y-channel from the corresponding normalized Y-channel to obtain the specular features,

concatenating the specular features and the texture-based feature vector to obtain a final feature vector; and

based on the final feature vector, ascertaining whether the facial image is of a facial substitute.

2. The system as claimed in claim 1 , wherein the image analysis module to detect the facial portion of the user based on one of edge detection techniques and a support vector machine classifier.

3. The system as claimed in claim 1 , wherein to derive the facial image from the processed captured image, the image analysis module is to:

cropping a portion of the captured image bearing the facial portion to obtain the facial image;

detecting pair of eyes within the facial image;

aligning the facial image with respect to a notional axis; and

resizing the facial image to a predefined size.

4. The system as claimed in claim 1 , wherein for determining the texture-based feature vector, the image analysis module is to:

converting the facial image to grayscale facial image; and

obtaining the texture-based feature vector based on applying a local binary pattern (LBP) function onto the grayscale facial image.

5. The system as claimed in claim 4 , wherein the applying the local binary pattern function further comprises:

for a cell comprising a predefined number of pixels of the facial image, applying the LBP function;

for a given cell, obtaining an LBP histogram based on the LBP function;

normalizing the LBP histogram; and

consolidating LBP histogram for all cells of the facial image to obtain the texture-based feature vector.

6. The system as claimed in claim 1 , wherein the ascertaining whether the facial image is of a facial substitute based on the specular features and the texture-based feature vector is based on a support vector classifier applied onto the final feature vector.

7. The system as claimed in claim 1 , wherein the captured image comprises one of a still image and a video stream.

8. A method comprising:

obtaining a captured image, wherein the captured image comprises a facial portion of a user;

processing the captured image to detect the facial portion of the user within the captured image;

generated a facial image from the processed captured image, wherein the facial image comprises the facial portion of the user;

for the generated facial image, determining a set of specular features and texture-based feature vector, wherein determining the set of specular features comprises:

converting the facial image to a YUV color space to obtain a converted facial image; and

processing the Y-channel of converted facial image to obtain the specular features, wherein the processing of the Y-channel further comprises:

normalizing the Y-channel;

performing a histogram normalization of the normalized Y-channel;

further performing an intensity transformation on the normalized Y-channel to obtain a transformed Y-channel; and

subtracting the transformed Y-channel from the corresponding normalized Y-channel to obtain the specular features,

concatenating the specular features and the texture-based feature vector to obtain a final set of feature vector; and

ascertaining whether the facial image corresponds to an actual face of the user based on the final set of feature vector.

9. The method as claimed in claim 8 , wherein the captured image in an RGB format and is one of a still image and a video stream.

10. The method as claimed in claim 8 , wherein generating the facial image comprises:

cropping a portion of the captured image bearing the facial portion to obtain the facial image;

aligning the facial image with respect to a notional axis based on detecting an angle of a notional line between eyes within the facial image with respect to the notional axis; and

resizing the facial image to a predefined size.

11. The method as claimed in claim 8 , wherein for determining the texture-based feature vector, the method further comprises:

converting the facial image to grayscale facial image; and

obtaining the texture-based feature vector based on applying a local binary pattern (LBP) function onto the grayscale facial image.

12. The method as claimed in claim 8 , wherein the ascertaining whether the facial image corresponds to an actual face of the user based on the final set of feature vector is based on a support vector classifier.