Systems And Methods For Providing Personal Video Services

Abstract

Systems and methods for processing video are provided. Video compression schemes are provided to reduce the number of bits required to store and transmit digital media in video conferencing or videoblogging applications. A photorealistic avatar representation of a video conference participant is created. The avatar representation can be based on portions of a video stream that depict the conference participant. A face detector is used to identify, track and classify the face. Object models including density, structure, deformation, appearance and illumination models are created based on the detected face. An object based video compression algorithm, which uses machine learning face detection techniques, creates the photorealistic avatar representation from parameters derived from the density, structure, deformation, appearance and illumination models.

1 . A method of video conferencing, the method comprising the steps of:
 detecting a human face of a video conference participant depicted in portions of a video stream;   creating by explicitly modeling one or more explicit object models to model the face of the video conference participant;   generating one or more implicit object models relative to parameters obtained from the explicit object models to facilitate creation of a compact encoding of the video conference participant's face; and   using the implicit object models, creating a photorealistic avatar representation of the video conference participant.   
     
     
         2 . A method of video conferencing, the method comprising the steps of:
 detecting a human face of a video conference participant depicted in portions of a video stream;   creating by explicitly modeling one or more explicit object models to model the face of the video conference participant;   generating one or more implicit object models relative to parameters obtained from the explicit object models to facilitate creation of a compact encoding of the video conference participant's face; and   using the implicit object models, creating a photorealistic avatar representation of the video conference participant.   
     
     
         3 . A method for providing video conferencing as in  claim 2  wherein the face of the video conference participant is detected and tracked using a Viola/Jones face detection algorithm. 
     
     
         4 . A method for providing video conferencing as in  claim 2  wherein the implicit object models provide an implicit representation of the face of the video conference participant. 
     
     
         5 . A method for providing video conferencing as in  claim 4  wherein the implicit representation of the video conference participant is a simulated representation of the face of the video conference participant. 
     
     
         6 . A method for providing video conferencing as in  claim 4  wherein the detecting and tracking comprise using a Viola/Jones face detection algorithm further includes the steps of:
 identifying corresponding elements of at least one object associated with the face in two or more video frames from the video stream; and 
 tracking and classifying the corresponding elements to identify relationships between the corresponding elements based on previously calibrated and modeled faces. 
 
     
     
         7 . A method for providing video conferencing as in  claim 1  wherein the explicit object models include one or more object models for structure, deformation, pose, motion, illumination, and appearance. 
     
     
         8 . A method for providing video conferencing as in  claim 1  wherein the implicit object models are configured using parameters obtained from the explicit object models, such that the explicit object model parameters are used as a ground truth for estimating portions of the video stream with the implicit object models. 
     
     
         9 . A method for providing video conferencing as in  claim 8  wherein the explicit object model parameters are used to define expectations about how lighting interacts with the structure of the face of the video conference participant. 
     
     
         10 . A method for providing video conferencing as in  claim 8  wherein the explicit object model parameters are used to limit a search space to the face or portions thereof for the implicit object modeling. 
     
     
         11 . A method for providing video conferencing as in  claim 1  further includes periodically checking to determine whether the implicit object modeling is working optimally. 
     
     
         12 . A method for providing video conferencing as in  claim 11  wherein periodically checking to determine whether the implicit object modeling is working optimally further includes determining that the implicit object models, which are used to create the photorealistic avatar representation, are working optimally by:
 determining that reprojection error is low in the photorealistic avatar representation; and 
 determining that there is a significant amount of motion in the photorealistic avatar representation. 
 
     
     
         13 . A method for providing video conferencing as in  claim 11  wherein the determination that the implicit object modeling is working optimally causes subsequent instances of the photorealistic avatar representation of the conference participant to be created without relying on the step of detecting a human face in the portions of the video stream. 
     
     
         14 . A method for providing video conferencing as in  claim 11  wherein determining that the implicit object modeling is not working optimally by:
 determining that processing of the photorealistic avatar representation uses a disproportional amount of transmission bandwidth; or 
 determining that the implicit object modeling is not working optimally if reprojection error is high. 
 
     
     
         15 . A method for providing video conferencing as in  claim 11  further includes responding to the determination that the implicit object modeling is not working by processing the step of detecting a human face of a video conference participant; and
 in response to detecting a human face, searching for existing calibration information for the detected human face. 
 
     
     
         16 . A method for providing video conferencing as in  claim 15  wherein if a human face is undetectable, using a Viola-Jones face detector to facilitate detection. 
     
     
         17 . A method for providing video conferencing as in  claim 1  wherein creating a photorealistic avatar representation of the video conference participant further includes enabling the video conference participant to adjust a gaze of their respective photorealistic avatar representation. 
     
     
         18 . A method for providing video conferencing as in  claim 1  wherein the gaze adjustment enables configuration of the gaze of the photorealistic avatar representation causes eyes of the photorealistic avatar representation to focus directly in the direction of a video camera. 
     
     
         19 . A computer program product for facilitating video conferencing, the computer program product being embodied on a computer-readable medium and comprising code configured so as when executed on a transmitting terminal to perform operations of:
 creating by explicitly modeling one or more explicit object models to model a detected face of a video conference participant;   generating one or more implicit object models relative to parameters obtained from the explicit object models to facilitate creation of a compact encoding of the video conference participant's face; and   using the implicit object models, creating a photorealistic avatar representation of the video conference participant.   
     
     
         20 . A video conferencing system comprising:
 a face detector configured to detect a face of a video conference participant in a video stream;   a calibrator configured to generate a calibration model calibrating the face of the video conference participant;   an explicit object modeler configured to generate one or more explicit object models, in combination with the calibrator and face detector, the explicit object models modeling portions of the video stream depicting the face of the video conference participant based on the calibration model;   an implicit object modeler configured to build one or more implicit object models relative to parameters from the explicit object models to facilitate creation of a compact encoding of the participant's face; and   the system operable to generate a photorealistic avatar representation of the video conference participant from the implicit models.