Over-parameterized variational optical flow method
First Claim
1. A method of determining the optical flow vector of a plurality of image pixels between two consecutive images in a group of images, the method comprising the steps of:
- (i) creating multiple sets of said group of images wherein each set contains a decreasing number of pixels per image;
(ii) defining an optical flow motion model applicable to the group of images in terms of a set of basis functions;
(iii) representing the optical flow vector of each image pixel by different coefficients of the same optical flow motion model;
(iv) starting with the set of group of images with the lowest number of pixels per image, determining respective over-parameterized optical flow vectors at the pixels responsively to motion between the images in the group by computing respective regularized coefficients of the basis functions at each of the pixels;
(v) interpolating the optical flow vectors determined in step (iv) to the next higher resolution set of group of images and refining the estimation using the over-parameterized optical flow representation; and
(vi) repeating step (v) until the highest resolution set of group of images is reached.
2 Assignments
0 Petitions
Accused Products
Abstract
An optical flow estimation process based on a spatio-temporal model with varying coefficients multiplying a set of basis functions at each pixel. The benefit of over-parameterization becomes evident in the smoothness term, which instead of directly penalizing for changes in the optic flow, accumulates a cost of deviating from the assumed optic flow model. The optical flow field is represented by a general space-time model comprising a selected set of basis functions. The optical flow parameters are computed at each pixel in terms of coefficients of the basis functions. The model is thus highly over-parameterized, and regularization is applied at the level of the coefficients, rather than the model itself. As a result, the actual optical flow in the group of images is represented more accurately than in methods that are known in the art.
-
Citations
20 Claims
-
1. A method of determining the optical flow vector of a plurality of image pixels between two consecutive images in a group of images, the method comprising the steps of:
-
(i) creating multiple sets of said group of images wherein each set contains a decreasing number of pixels per image; (ii) defining an optical flow motion model applicable to the group of images in terms of a set of basis functions; (iii) representing the optical flow vector of each image pixel by different coefficients of the same optical flow motion model; (iv) starting with the set of group of images with the lowest number of pixels per image, determining respective over-parameterized optical flow vectors at the pixels responsively to motion between the images in the group by computing respective regularized coefficients of the basis functions at each of the pixels; (v) interpolating the optical flow vectors determined in step (iv) to the next higher resolution set of group of images and refining the estimation using the over-parameterized optical flow representation; and (vi) repeating step (v) until the highest resolution set of group of images is reached. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
-
-
11. A non-transitory computer-readable medium encoded with a program module that determines the optical flow vector of a plurality of image pixels between two consecutive images in a group of images, by:
-
(i) creating multiple sets of said group of images wherein each set contains a decreasing number of pixels per image; (ii) defining an optical flow motion model applicable to the group of images in terms of a set of basis functions; (iii) representing the optical flow vector of each image pixel by different coefficients of the same optical flow motion model; (iv) starting with the set of group of images with the lowest number of pixels per image, determining respective over-parameterized optical flow vectors at the pixels responsively to motion between the images in the group by computing respective regularized coefficients of the basis functions at each of the pixels; (v) interpolating the optical flow vectors determined in step (iv) to the next higher resolution set of group of images and refining the estimation using the over-parameterized optical flow representation; and (vi) repeating step (v) until the highest resolution set of group of images is reached. - View Dependent Claims (12, 13, 14, 15, 16, 17, 18, 19, 20)
-
Specification