Spatial clustering of pixels of a multispectral image
View Patent ↗A method and system for clustering the pixels of a multispectral image is provided. A clustering system computes a maximum spectral similarity score for each pixel that indicates the similarity between that pixel and the most similar neighboring. To determine the maximum similarity score for a pixel, the clustering system generates a similarity score between that pixel and each of its neighboring pixels and then selects the similarity score that represents the highest similarity as the maximum similarity score. The clustering system may apply a filtering criterion based on the maximum similarity score so that pixels with similarity scores below a minimum threshold are not clustered. The clustering system changes the current pixel values of the pixels in a cluster based on an averaging of the original pixel values of the pixels in the cluster.
1. A method in a computing device for spatial clustering of pixels of a multispectral image, the method comprising:
determining a maximum similarity score for each pixel by
computing a similarity score between that pixel and each of its neighboring pixels, the similarity score based on pixel values; and
setting the maximum similarity score for that pixel to the maximum of the computed similarity scores for that pixel; and
clustering the pixels into clusters by, when a filtering criterion is passed for a pixel,
combining into the same cluster that pixel and the neighboring pixel with the maximum similarity score; and
setting the pixel values of the pixels in the cluster to an average of the original pixel values of the pixels that are in the cluster; and
repeating iterations of the determining and the clustering until a termination criterion is satisfied.
2. The method of claim 1 wherein the similarity scores are a cosine similarity.
3. The method of claim 1 wherein the similarity scores are based on an arccosine of a normalized dot product of intensity vectors of the pixels.
4. The method of claim 1 wherein the similarity scores are based on an inverse of an arccosine of a normalized dot product of intensity vectors of the pixels.
5. The method of claim 1 wherein the filtering criterion is a per-iteration minimum threshold for the maximum similarity score.
6. The method of claim 1 wherein the filtering criterion is an overall minimum threshold for the maximum similarity score.
7. The method of claim 1 wherein when a pixel is already in a cluster of pixels, all pixels of the cluster are included in the combining.
8. The method of claim 1 wherein the filtering criterion is a first-iteration minimum threshold.
9. The method of claim 1 wherein the filtering criterion is a maximum cluster size.
10. A computer-readable memory device containing computer-executable instructions for controlling a computing device to cluster contiguous pixels of a multispectral image, by a method comprising:
computing a maximum similarity score for each pixel that indicates the similarity of the most similar neighboring pixel based on a comparison of pixel values of the pixels;
clustering neighboring pixels when a pixel has a maximum similarity score that passes a filtering criterion wherein the pixel values of the pixels in the cluster are set based on an averaging of the pixel values of the pixels in the cluster; and
repeating iterations of the computing and clustering until a termination criterion is satisfied.
11. The computer-readable memory device of claim 10 wherein the termination criterion is based on a number of iterations.
12. The computer-readable memory device of claim 10 wherein the termination criterion is based on changes to the clusters during the current iteration.
13. The computer-readable memory device of claim 10 wherein the similarity is based on a cosine-based similarity.
14. The computer-readable memory device of claim 10 wherein the filtering criterion is per-iteration minimum threshold.
15. The computer-readable memory device of claim 10 wherein when a pixel is already in a cluster of pixels, all pixels of the cluster are included in any clustering of that pixel.
16. The computer-readable memory device of claim 10 wherein the filtering criterion is first-iteration minimum threshold.
17. The computer-readable memory device of claim 10 wherein the filtering criterion is per-cluster similarity threshold.
18. A computing device for clustering pixels of a multispectral image, comprising:
a memory storing computer-executable instructions of:
a compute similarities component that computes a maximum similarity score for each pixel indicating its similarity to the most similar neighboring pixel based on a comparison of pixel values of the pixels;
a grow clusters component that clusters neighboring pixels when a pixel has a maximum similarity score that passes a filtering criterion wherein the pixel values of pixels in the cluster are set based on an averaging of the pixel values of the pixels in the cluster; and
a component that repeatedly invokes the compute similarities component and the grow clusters component until a termination criterion is satisfied; and
a processor that executes the computer-executable instructions stored in the memory.
19. The computing device of claim 18 wherein the filtering criterion is a per-iteration minimum threshold.
20. The computing device of claim 18 wherein the filtering criterion is a first-iteration minimum threshold.
21. The method of claim 1 wherein the computing of the similarity score weights frequencies of interest in the multispectral image more heavily than other frequencies.
22. The method of claim 1 wherein pixels are processed in a row-major ordering.