| 摘要 |
Techniques to determine image similarity. In one embodiment, a small hash value associated with an image is determined to fall within a bucket. If an existing cluster center is associated with the bucket, the distance between the small hash value and the cluster center is determined. If the distance from the small hash value to the cluster center is greater than a threshold distance, then the image is considered to be different from other images associated with the cluster center. A new cluster center is then created and associated with the bucket, and the image is associated with the new cluster center. More than one cluster center may be associated with the bucket when images having small hash values falling within the bucket are not considered to be similar. In another embodiment, once a bucket is identified, the distance between a large hash value, instead of a small hash value, and a cluster center may be determined to assess image similarity. |
