Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1999-07-02
2003-12-16
Le, Vu (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
Reexamination Certificate
active
06665342
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to the field of digital video and/or image processing. More specifically, the invention relates to producing one or more still image representations of a motion video.
BACKGROUND OF THE INVENTION
A still image representation of a simple motion video provides a visual record of an objects' motion. Depending on the specification of a timing parameter, the still image depicts the displacements of the object or objects over time. This image representation of the motion inherent in the video sequence can be used to visually inspect, measure, and analyze the event.
Using physical devices, conventional film photographers have been able to create imagery of moving objects using several techniques. For slow movement, including time-lapse, photographs are created by automatically opening the shutter using a timer and, if the action is to be recorded on a single photograph, exposing the film multiple times. For outdoor images which may have variable lighting, a powerful strobe light may also be used to conceal the natural variation in the lighting. In high-speed photography, either a very fast shutter and/or short brilliant flashes of light from a strobe lamp are used in order to capture the action while providing adequate light. Various mechanical and electronic devices are used to record the action on film, such as rotating mirrors and prisms, rotating shutters with various slits, film spools, and timers for synchronizing flashes and shutters. In general, to shoot a conventional film photograph showing the movement of objects over time, requires specific equipment such as strobe lights, timers, and automatic shutters, and needs to be done at the scene of the action of interest.
In the field of digital video, procedures for mixing and performing luminance and chroma (chromatic) keying are used to overlay text or graphics, fade from one sequence into another, and combine the foreground of one scene with the background of another. In particular, keying, is used to substitute the background signal in portions of the foreground signal based on some attribute of the video signal. In chroma keying, also referred to as blue-screen, an easily segmented background (often this is blue, hence the term “blue-screen”) is used so that the action in the foreground can be extracted by a simple thresholding of the chrominance value at each point and overlaid on a different background.
In digital background subtraction, more sophisticated techniques are used to segment out the background. Three main approaches are taken. In the first class of methods, statistical texture properties of the background are observed over an extended period of time. This requires additional footage and preprocessing. The second class of methods is based upon geometry and relies on the assumption that the geometry of the background remains stationary over time. For these methods, specific hardware configurations such as stereo or multiple cameras are necessary. The third class of the methods, of which this process is most closely related, is based upon presuming that the background is stationary or at most slowing changing. These methods are often devised as preliminary steps to representing and recognizing actions of a specific nature, such as human gestures or facial expressions or tracking an object over time. Some of the prior art is included in, e.g., “Motion-based recognition: A survey. Image and Vision Computing”, by Cedras, C., and Shah, M.“Image and Vision Computing” 13(2):129-155, March '95, which is herein incorporated by reference in its entirety. These methods create either representations of specific motions in the case of motion recognition, or trajectories of objects in the case of visual tracking. The motion representations include three dimensional models, sequences of features or properties, or characteristics of temporal and spatial changes. Since recognition requires a noisy and large search space, these methods are designed to create representations which are easily indexed, compact, and invariant to small variations. For visual tracking, correspondence is necessary, either at the image level, to perform optic flow, region segmentation and matching, or at the feature level, to match features over time. In general, this is a much more complex task involving image preprocessing for segmentation or feature detection, geometric and temporal constraints to limit search, and matching techniques, such as image correlation or warping.
PROBLEMS WITH THE PRIOR ART
Some prior art requires strobe lights, timers, automatic shutters, and other analog equipment. It is also necessary to acquire the photographs at the scene, during the event, and can only be done with a single setting of parameters. Strobe lights are triggered at fixed time intervals and shutter speeds are similarly set to specific speeds. Once photographs are acquired, the only way to alter these settings is to recapture a similar event with a new set of parameters.
Some prior art can only be used with a fixed physical background such as a uniformly lit blue screen. With these simple digital techniques, the background needs to be trivially segmented, i.e., above or below a luminance or chromatic threshold.
Some prior art was developed for the purposes of recognizing specific actions or visually tracking the motion of objects. Statistical methods require preprocessing of footage containing only background information. Geometric methods require two or more cameras and/or perform a computationally costly and inherently noisy correspondence mapping. For motion recognition or tracking methods which assume a stationary background, these techniques generally perform some form of correspondence; they are developed to perform higher level functions and as a consequence these methods do not create a general visual representation for inspecting and measuring motion, and do not directly address or exploit the properties of the background/foreground segmentation with regard to creating a visual depiction of the motion history.
OBJECTS OF THE INVENTION
An object of this invention is an improved system and method for visualizing motion in video.
An object of this invention is an improved system and method for inspecting, analyzing and measuring motion in video.
SUMMARY OF THE INVENTION
This invention is a system and method for updating a strobe photo with one or more relevant motions at one or more photo locations in the strobe photo. The system is a computer system that stores and/or processes one or more well known digital video streams. (The stream has a plurality of still frames in a sequence called an ordered sequence.) A strobe process, executed by the system first initializes a motion history record with one or more motion statistics, a segmentation mask, an attribute, and an attribute sensitivity. Two of the still frames are iteratively selected in a sampled order. An attribute differencing step is performed that determines the difference of the attribute of the selected frames using a frame to frame comparison to determine an attribute difference at one or more frame locations in the frames and that further determines if the attribute difference from one frame to the other frame exceeds the attribute sensitivity. The attribute difference is then stored in the motion history record along with the motion statistics of the attribute difference. If the attribute difference exceeds the attribute sensitivity when the segmentation mask has not yet been updated, a relevant motion is determined. If the segmentation mask has been updated and the attribute difference exceeds one or more of the motion statistics within a statistical confidence level, a relevant motion is also determined. If these conditions are met (to determine a relevant motion), the segmentation mask, at one or more mask locations corresponding to the frame locations, is updated. If the segmentation mask is updated, a strobe photo is also updated with the relevant motion at one or more photo locations corresponding to the mask and frame locations
Brown Lisa Marie
Crayne Susan
International Business Machines - Corporation
Le Vu
Percello Louis J.
Senfi Behrooz
LandOfFree
System and method for producing a still image representation... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for producing a still image representation..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for producing a still image representation... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3119274