Systems and methods of detecting significant faces in video streams

摘要

Systems and methods of processing video streams are described. A face is detected in a video stream. The face is tracked to determine a video clip associated with one of a plurality of individuals. The video segment is assigned to a group of video clips based on the associated individual. A significant face is detected in the group of video clips when the detected face meets one or more significance criteria. The significance criteria describes a face-frame characteristic. A representation of the significant face is displayed in association with a representation of the group of video clips. The order of the significance criteria is adjusted through a user interface.

主权项

1. A method of detecting significant faces in video streams comprising:
detecting at least one face in a video stream; tracking the at least one face to determine at least one video clip associated with one of a plurality of individuals; assigning the at least one video clip to a group of video clips based on the associated one individual; detecting at least one significant face in the group of video clips when the detected face meets one or more significance criteria, wherein the significance criteria is associated with a face-frame characteristic; displaying a representation of the at least one significant face in association with a representation of the group of video clips; and adjusting an order of the one or more significance criteria through a user interface, wherein the tracking the at least one face is performed using backward tracking or motion object tracking, wherein the backward tracking comprises:
noting a frame in which a greater portion of the detected face comes into view; andbackward tracking the detected face by tracking an object in those frames in the video stream that are earlier in time than the noted frame to find the first occurrence of the detected face in the video stream.