Television – Object tracking
Patent
1999-01-19
2000-08-08
Tung, Bryan
Television
Object tracking
348584, H04N 718
Patent
active
061009253
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a system and method for tracking image frames for inserting realistic indicia into video images.
2. Description of Related Art
Electronic devices for inserting electronic images into live video signals, such as described in U.S. Pat. No. 5,264,933 by Rosser, et al., have been developed and used for the purpose of inserting advertising and other indicia into broadcast events, primarily sports events. These devices are capable of seamlessly and realistically incorporating logos or other indicia into the original video in real time, even as the original scene is zoomed, panned, or otherwise altered in size or perspective. Other examples include U.S. Pat. No. 5,488,675 issued to Hanna and U.S. Pat. No. 5,491,517 issued to Kreitman, et al.
Making the inserted indicia look as if it is actually in the scene is an important but difficult aspect of implementing the technology. A troublesome aspect is that the eye of the average viewer is very sensitive to small changes in the relative position of objects from field to field. Experimentally, instances have been found where relative motion of an inserted logo by as little as one tenth of one pixel of an NTSC television image is perceptible to a viewer. Placing, and consistently maintaining to a high precision, an inserted indicia in a broadcast environment is crucial in making video insertion technology commercially viable. A broadcast environment includes image noise, the presence of sudden rapid camera motion, the sporadic occurrence of moving objects which may obscure a considerable fraction of the image, distortions in the image due to lens characteristics and changing light levels, induced either by natural conditions or by operator adjustment, and the vertical interlacing of television signals.
In the prior art, the automatic tracking of image motion has generally been performed by two different methods.
The first method utilizes pattern recognition of the frames and examines the image itself and either follows known landmarks in the video scene, using correlation or difference techniques, or calculates motion using well known techniques of optical flow. See, Horn, B. K. P. and Schunck, B. G., "Determining Optical Flow", Artificial Intelligence, pp 185-203 (1981). Landmarks may be transient or permanent and may be a natural part of the scene or introduced artificially. A change in shape and pose of the landmarks is measured and used to insert the required indicia.
The second method, described, for instance, in U.S. Pat. No. 4,084,184 issued to D. W. Crain, uses sensors placed on the camera to provide focal distance, bearing and elevation information. These sensors exist to provide similar landmark positional data within a given camera's field of view.
In the pattern recognition type of image insertion systems developed by Rosser et al., for instance, the system has two distinct modes. First is the search mode wherein each new frame of live video is searched in order to detect and verify a particular target image. Second is the tracking mode, in which the system knows that in the previous frame of video the target image was present. The system further knows the location and orientation of that previous frame with respect to some pre-defined reference coordinate system. The target image locations are tracked and updated with respect to the pre-defined reference coordinate system.
The search mode encompasses pattern recognition techniques to identify certain images. Obtaining positional data via pattern recognition, as opposed to using camera sensors, provides significant system flexibility because it allows live video insertion systems to make an insertion at any point in the video broadcast chain. For instance, actual insertion can be performed at a central site which receives different video feeds from stadiums or arenas around the country or world. The various feeds can be received via satellite or cable or any other means known in the art. Once the insertion is added, the video f
REFERENCES:
patent: 5264933 (1993-11-01), Rosser
patent: 5436672 (1995-07-01), Medioni
patent: 5488675 (1996-01-01), Hanna
patent: 5491517 (1996-02-01), Kreitman
patent: 5566251 (1996-10-01), Hanna
DiCicco Darrell S.
Gong Ximin
Jeffers James L.
Kennedy, Jr. Howard J.
Rosser Roy J.
Princeton Video Image, Inc.
Tung Bryan
LandOfFree
Image insertion in video streams using a combination of physical does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image insertion in video streams using a combination of physical, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image insertion in video streams using a combination of physical will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1155246