Videoconference Eye Contact via Gaze Detection and Image Processing

The patent creates a virtual copy of the user's eye region through image processing. Instead of using multiple physical cameras, the system captures the user's face with a single camera, extracts the eye region, and generates a processed version that simulates direct eye contact appearance. This virtual copy is then displayed to remote users, achieving the eye contact effect without additional hardware.

The patent replaces the mechanical/optical system of multiple cameras and special displays with a software-based image processing system. The complex hardware arrangement (multiple cameras positioned at specific angles, special display configurations) is substituted with algorithms that detect eye direction and manipulate the captured image to create the appearance of eye contact, thereby eliminating the need for complex mechanical setups.

The system creates a processed copy of the user's eye region that maintains the appearance of looking at the camera while actually displaying the user looking in the direction of the screen. This allows the user to simultaneously view the other participant and maintain eye contact, as the processed image shows eyes directed toward the camera position rather than the screen position.

Instead of having the user look away from the camera to see the screen (which breaks eye contact), the system inverts the approach by processing the image to show the user's eyes appearing to look at the camera even when they are looking at the screen. This reverses the traditional problem-solution dynamic, allowing both screen viewing and eye contact to coexist.