Using directional weighting functions while computing optical flow through belief propagation
AssigneeAdobe Systems Incorporated
InventorMark A. Ruzon
Life of this patent
Estimating the term…
Abstract
One embodiment of the present invention provides a system that uses directional weighting functions while computing optical flow for an image through belief propagation. The system operates by computing local velocity estimates for points in the image. The system also computes directional weighting functions for points in the image, wherein the directional weighting functions facilitate propagating information in a manner which can vary with direction. The system subsequently uses these directional weighting functions during a belief propagation process, which propagates the local velocity estimates between points in the image, and which computes an optical flow field for points in the image.
Claims25
A method for computing optical flow for an image through belief propagation, comprising:
- computing local velocity estimates for points in the image;
- computing directional weighting functions for points in the image, wherein the directional weighting functions facilitate propagating information in a manner which can vary with direction;
- and
- using the directional weighting functions during a belief propagation process, which propagates the local velocity estimates between points in the image, and which computes an optical flow field for points in the image.
A computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for computing optical flow for an image through belief propagation, the method comprising:
- computing local velocity estimates for points in the image;
- computing directional weighting functions for points in the image, wherein the directional weighting functions facilitate propagating information in a manner which can vary with direction;
- and
- using the directional weighting functions during a belief propagation process, which propagates the local velocity estimates between points in the image, and which computes an optical flow field for points in the image.
An apparatus that computes optical flow for an image through belief propagation, comprising:
- a velocity computing mechanism configured to compute local velocity estimates for points in the image;
- a directional weighting function mechanism configured to compute directional weighting functions for points in the image, wherein the directional weighting functions facilitate propagating information in a manner which can vary with direction;
- and
- a belief propagation mechanism which propagates the local velocity estimates between points in the image, and which computes an optical flow field for points in the image, wherein the belief propagation mechanism is configured to use the directional weighting functions during a belief propagation process.
A method for computing an apparent motion for one or more objects in an image in a sequence of images, comprising:
- examining the image in the sequence of images;
- computing local velocity estimates for points in the image;
- computing directional weighting functions for points in the image, wherein the directional weighting functions facilitate propagating velocity information in a manner which can vary with direction;
- using the directional weighting functions during a belief propagation process, which propagates the local velocity estimates between points in the image, and which computes an optical flow field for points in the image;
- and
- using the computed optical flow field for points in the image to compute an apparent motion for one or more objects in the image.
Dependencies are inferred from the claim text.