User recognition and gaze tracking in a video system

A system for tracking a gaze of a user, the system comprising:

• one or more processors;
• and
• one or more machine-readable non-transitory storage mediums that includes instructions configured to cause one or more processors to perform operations including:
• receiving image data that includes a plurality of eye landmarks that identify a perimeter of a user's eye;
• applying dynamic thresholding to the image data including:
• determining a lowest pixel intensity value of the image data;
• determining a highest pixel intensity value of the image data;
• selecting a threshold value for the image data that is between the lowest and highest pixel intensity values;
• reassigning pixel values below the threshold value to a minimum intensity value;
• and
• reassigning pixel values at or above the threshold value to a maximum intensity value;
• determining a convex hull based on the dynamically thresholded image data;
• computing and fitting an ellipse along a boundary of the convex hull, the ellipse corresponding to an iris of the user's eye;
• updating the image data with the computed ellipse;
• computing a refined ellipse by reapplying the dynamic thresholding and redetermining the convex hull using the updated image data;
• and
• determining a gaze direction of the user based in part on a position of the refined ellipse relative to the perimeter of the user's eye.

A non-transitory computer-program product tangibly embodied in a machine-readable non-transitory storage medium that includes instructions configured to cause one or more processors to perform operations including:

• receiving image data that includes a plurality of eye landmarks that identify a perimeter of a user's eye;
• applying dynamic thresholding to the image data including:
• determining a lowest pixel intensity value of the image data;
• determining a highest pixel intensity value of the image data;
• selecting a threshold value for the image data that is between the lowest and highest pixel intensity values;
• reassigning pixel values below the threshold value to a minimum intensity value;
• and
• reassigning pixel values at or above the threshold value to a maximum intensity value;
• determining a convex hull based on the dynamically thresholded and image data;
• computing and fitting an ellipse along a boundary of the convex hull, the ellipse corresponding to an iris of the user's eye;
• updating the image data with the computed ellipse;
• computing a refined ellipse by reapplying the dynamic thresholding and determining the convex hull using the updated image data;
• and
• determining a gaze direction of the user based in part on a position of the refined ellipse relative to the perimeter of the user's eye.