Representing image patches for matching
View Patent ↗Processes such as image matching can be improved by utilizing a Polar Accumulated Signed Gradient (PASG), a feature descriptor vector that utilizes polar coordinates. Such an approach is efficient because it avoids the costly image rotation and resampling of conventional processes. Further, the gradients can be sampled over different regions in order to achieve effective weighting without analyzing all gradients and applying a weighting factor. Such efficiencies enable image matching to be performed on devices with limited computing capacity, such as smart phones and tablet computers.
1. A computer-implemented method of generating a histogram for an image, comprising:
under control of one or more computing systems configured with executable instructions, determining a set of keypoints for an image;
analyzing a circular region of the image around each of the set of keypoints;
determining a dominant orientation of each of the circular regions based at least in part upon a plurality of gradients of each of the circular regions;
positioning a polar spatial binning layout to be centered on each of the set of keypoints, each of the polar spatial binning layouts being rotated relative to the dominant orientation;
extracting a set of gradients from pixel locations within the bins of each polar spatial binning layout, the extracted gradients being rotated relative to the dominant orientation; and
generating a histogram by accumulating the gradients into positions based at least in part on the bin assigned to each gradient and the direction of the gradient.
2. The computer-implemented method of claim 1 , wherein each polar spatial binning layout contains a central bin and a plurality of concentric annular bins.
3. The computer-implemented method of claim 2 , further comprising:
determining a number of gradients to be processed for each bin of the polar spatial binning layout in order to adjust a relative contribution of the bins.
4. The computer-implemented method of claim 3 , wherein the relative contribution is an even contribution, a contribution that decreases linearly with increasing distance from the keypoint, or a contribution that decreases logarithmically with increasing distance from the keypoint.
5. The computer-implemented method of claim 3 , wherein a user is able to adjust the number of gradients to be processed in order to adjust at least one of speed or accuracy.
6. The computer-implemented method of claim 1 , further comprising:
determining a set of candidate keypoints in the image by locating local minima or maxima among neighboring pixels in the image; and
processing the set of candidate keypoints to select the set of keypoints.
7. The computer-implemented method of claim 1 , wherein generating the histogram does not require a rotation or resampling of the image.
8. The computer-implemented method of claim 1 , wherein each gradient is assigned to a bin based at least in part upon the dominant orientation of the corresponding polar spatial binning layout.
9. The computer-implemented method of claim 1 , further comprising:
normalizing the histogram.
10. A non-transitory computer-readable storage medium storing instructions that, when executed by at least one processor of a computing device, cause the computing device to:
determine a set of keypoints for an image;
determine a dominant orientation of each of the circular regions by analyzing a region of the image around each of the set of keypoints;
position a polar spatial binning layout to be centered on each of the set of keypoints, each of the polar spatial binning layouts being rotated relative to the determined dominant orientation;
extract a set of gradients from locations within the bins of each polar spatial binning layout, the extracted gradients being rotated relative to the dominant orientation; and
generate a histogram by accumulating the gradients into positions based at least in part on the bin assigned to each gradient and the direction of the gradient.
11. The non-transitory computer-readable storage medium of claim 10 , wherein generating the histogram does not require a rotation or resampling of the image.
12. The non-transitory computer-readable storage medium of claim 10 , wherein each gradient is assigned to a bin based at least in part upon the dominant orientation of the corresponding polar spatial binning layout.
13. The non-transitory computer-readable storage medium of claim 10 , wherein the instructions when executed further cause the computing device to:
determine a set of candidate keypoints in the image by comparing at least a portion of pixels in the image to neighboring pixels in the image; and
process the set of candidate keypoints to determine the set of keypoints.
14. The non-transitory computer-readable storage medium of claim 10 , wherein the instructions when executed further cause the computing device to:
determine a number of gradients to be processed for each bin.