Apparatus and method of determining motion vectors assigned to image regions
View Patent ↗An image includes a plurality of image blocks each including a plurality of pixels and having an associated motion vector. The image is processed by selecting an image block and a corner region (E 1 x ) located in a corner of the selected image block to provide a selected corner region. A plurality of adjacent image blocks located adjacent to the selected corner region are then selected from the plurality of image blocks. The motion vectors assigned to the adjacent image blocks are then compared to determine if the motion vectors form a homogeneous vector field. When the motion vectors of the adjacent image blocks form a homogeneous vector field, a motion vector for the corner region is generated from the motion vectors of the adjacent image blocks. When the motion vectors of the adjacent image blocks do not form a homogeneous vector field, the motion vector for the corner region is set equal to the motion vector of the image block.
1. A method of determining motion vectors which are assigned to individual image regions of an image, wherein the image is subdivided into a number of rectangular image blocks, and a motion vector is assigned to each block of the number of rectangular image blocks, the method comprising:
defining, using a first image processing unit, a corner region located in a corner of a selected one of the number of rectangular image blocks;
determining, using a second image processing unit, a set of image blocks adjacent to the corner region;
determining, using a third image processing unit, whether the motion vectors assigned to the set of image blocks adjacent to the corner region form a homogeneous vector field, and in response to a determination that the motion vectors assigned to the set of image blocks adjacent to the corner region form a homogeneous vector field, generating, using a fourth image processing unit, a motion vector for the corner region from the motion vectors of the set of image blocks adjacent to the corner.
2. The method of claim 1 , comprising:
defining the motion vector of the selected one of the image blocks as the motion vector for the corner region whenever the motion vectors of the adjacent image blocks do not form the homogeneous vector field.
3. The method of claim 1 , where generating the motion vector for the corner region comprises averaging the motion vectors of the set of image blocks adjacent to the corner region.
4. The method of claim 3 , in which the motion vectors of the adjacent image blocks are weighted equally during averaging.
5. The method of claim 3 , in which the motion vector of one of the adjacent image blocks which is adjacent at only one corner to the corner region is given a higher weight during averaging than the motion vectors of the other image blocks adjacent to the corner region.
6. The method of claim 5 , in which the corner region comprises a corner pixel located in the corner of the selected image block.
7. The method of claim 6 , in which starting from the corner pixel the corner region comprises at least a first one pixel located in the horizontal direction of the selected image block adjacent to the corner pixel, and at least a second one pixel located in the vertical direction of the selected image block adjacent to the corner pixel.
8. The method of claim 5 , comprising:
defining, using a fifth image processing unit, a center region located in the center of the selected image block;
determining, using a sixth image processing unit, whether the motion vectors of at least five of the number of rectangular image blocks adjacent to the selected image block form a second homogeneous vector field; and
generating, using a seventh image processing unit, a motion vector for the center region from the motion vectors of the at least five adjacent image blocks whenever the motion vectors of these at least five adjacent image blocks form the second homogeneous vector field.
9. The method of claim 1 , where the presence of the second homogeneous vector field is determined when absolute values of difference vectors, which are formed from vector differences of the motion vectors of the adjacent image blocks, are smaller than a predetermined threshold value.
10. The method of claim 1 , where image blocks comprise four pixels horizontally and four pixels vertically.
11. An apparatus for determining motion vectors which are assigned to individual image regions of an image, wherein the image is subdivided into a number of rectangular image blocks, and a motion vector is assigned to each image block, comprising:
a unit that selects a corner region located in a corner of a selected one of the image blocks;
a unit that determines the image blocks adjacent to the corner region;
a unit that determines whether the motion vectors assigned to the image blocks adjacent to the corner region form a homogeneous vector field; and
a unit that generates a motion vector for the corner region from the motion vectors of the selected adjacent image blocks in response to a determination that the motion vectors of the adjacent image blocks form the homogeneous vector field.
12. A method of processing an image that includes a plurality of image blocks each including a plurality of pixels and having an associated motion vector, the method comprising:
selecting, using an image processing unit, an image block and a corner region located in a corner of the selected image block to provide a selected corner region;
determining, using the unit, a plurality of adjacent image blocks located adjacent to the selected corner region from the plurality of image blocks;
determining, using the unit, whether the motion vectors assigned to the adjacent image blocks form a homogeneous vector field; and
generating, using the unit, a motion vector for the corner region from the motion vectors of the adjacent image blocks in response to a determination that the motion vectors of the adjacent image blocks form a homogeneous vector field.
13. The method of claim 12 , where generating the motion vector for the selected corner region comprises averaging the motion vectors of the adjacent image blocks.
14. The method of claim 13 , where generating the motion vector for the selected corner region comprises applying a weighted averaging of the motion vectors of the adjacent image blocks.