Back to results
US 6,285,711 B1Granted
Block matching-based method for estimating motion fields and global affine motion parameters in digital video sequences
AssigneeSharp Laboratories of America, Inc.
InventorsKrishna Ratakonda, M. Ibrahim Sezan
Life of this patent
Estimating the term…
Abstract
A method of estimating a motion field includes estimating a first motion vector component from a source average vector and a target average vector; displacing the search area according to the first motion vector component; and estimating a second motion vector component from the displaced search area.
Claims17
A method of estimating motion fields comprising:
- estimating a first motion vector component in a first single direction from a source average vector and a target average vector;
- displacing the search area according to the first motion vector component;
- and
- estimating a second motion vector component in a second single direction at 90° to the first single direction from the displaced search area.
1 more element
A method of estimating motion fields comprising:
- estimating a first motion vector component in a first single direction from a source average vector and a target average vector;
- displacing the search area according to the first motion vector component;
- estimating a second motion vector component in a second single direction at 90° from the first single direction from the displaced search area;
- and
- spatially decimating two video images by a first predetermined decimation factor;
- Qn, deriving motion estimates;
- multiplying the derived motion estimates by Qp;
- aligning, as a function of the motion estimate multiplied by Qp, the original images decimating by a second predetermined decimation factor, Qn−p where Q is a given positive integer, n is an integer greater than zero, and p is an integer having a value between 1 and n.
5 more elements
A method of estimating motion fields comprising:
- estimating a first motion vector component in a first single direction from a source average vector and a target average vector;
- displacing the search area according to the first motion vector component;
- recentering the search area as a function of the previous iteration;
- estimating a second motion vector component in a second single direction at 90° to the first single direction from the displaced search area;
- spatially decimating, in a hierarchical scheme, two video images by a first predetermined decimation factor, Qn, deriving motion estimates;
- multiplying the derived motion estimates by Qp;
- aligning, as a function of the motion estimate multiplied by Qp, the original images decimating by a second predetermined decimation factor, Qn−p, where Q is a given positive integer, n is an integer greater than zero, and p is an integer having a value between 1 and n;
- and
- estimating global translation components by using a technique taken from the group of techniques consisting of an integral projection technique and quasi-projection technique;
- aligning the two images according to the global translation estimate of the preceding step;
- subdividing the image into small, non-overlapping blocks;
- selecting blocks whose content exhibit large variations;
- applying a quasi-projection matching technique to a selected image block to align the images to produce a motion vector;
- evaluating a matching error for a selected image block using the produced motion vector for the image block;
- counting the number of image blocks having matching errors below a predetermined threshold value T2 and storing motion vectors for such blocks;
- estimating global affine parameters if the number of image blocks having matching errors below T2 is greater than T3;
- estimating an affine parameter using the motion vectors of all image blocks having a matching error below T2.
14 more elements
Dependencies are inferred from the claim text.