Back to results
US 5,675,376 AGranted
Method for achieving eye-to-eye contact in a video-conferencing system
AssigneeLucent Technologies Inc.
InventorsRussell Lennart Andersson, Homer H. Chen
Life of this patent
Estimating the term…
Abstract
Methods for achieving eye-to-eye contact in a video-conferencing system use a controller to calculate eye contours, radius profiles, and positions from images stored in memory. Once the contours and radius profiles are generated, an eye image can be "shifted" to achieve eye-to-eye contact.
Claims26
A method for achieving eye-to-eye contact in a video-conferencing system comprising:
- controlling pixels in memory, each pixel representing part of an image comprising human eye positions and each having a corresponding intensity level, to achieve eye-to-eye contact, said controlling comprising,
- (1) first detecting the positions of each iris of each eye, said first detecting comprising instructing a controller to complete a spiral search of pixels stored in memory for each iris, categorizing the pixels searched, by intensity, as at least iris, near-iris or non-iris, calculating an initial pixel from said categorized iris intensities, instructing the controller to first scan the pixels in memory, including the categorized pixels, for each iris starting from the initial pixel of each iris until the controller locates pixels having intensity levels exceeding a first threshold over an average iris intensity level, the first scan comprising scanning a number of video lines of the image, further instructing the controller to halt said first scanning when said controller detects non-iris pixels, wherein an iris boundary for each iris is determined from said non-iris pixels, calculating a center for each iris from the iris boundary, and repeating said first detecting for each position of each iris,
- (2) second detecting the positions of each eyelid that covers each eye, said second detecting comprising instructing the controller to additionally scan pixels outside of the iris boundary of each eye in first directions, detecting pixel intensities corresponding to post-lid pixels, retracing said additional scanning in directions opposite to said first directions, detecting pixel intensities corresponding to last white pixels, said last white pixels indicating the beginning of each eyelid, instructing said controller to connect the last white pixels to generate a contour of each eye, and repeating said second detecting for each position of each eyelid, and
- (3) shifting positions of pixels stored in memory using the controller within each eyelid boundary of each eye to achieve eye-to-eye contact, and repeating said shifting for each position of each eye.
1 more element
A method for achieving eye-to-eye contact in a video-conferencing system comprising:
- controlling pixels in memory, each pixel representing part of an image comprising human eye positions and each having a corresponding intensity level, to achieve eye-to-eye contact, said controlling comprising,
- (1) instructing a controller to first scan pixels in memory and to determine an iris boundary and a center of each iris, for each position of each eye and repeating said first scan for each position of each iris,
- (2) instructing the controller to additionally scan pixels outside of the iris boundary of each eye and to generate a contour of each eye for each eye position from said additional scan and repeating said additional scan for each position of each eye, and
- (3) shifting positions of pixels using the controller within each eyelid boundary of each eye to achieve eye-to-eye contact, and repeating said shifting for each position of each eye.
1 more element
Dependencies are inferred from the claim text.