| 摘要 |
A search engine determines a set of other images that are similar to a user-selected image, and presents those other images to the user. In determining whether two images are sufficiently similar to each other to merit presentation of one, the search engine determines a Euclidean distance between separate feature vectors that are associated with each of the images. Each such vector indicates diverse types of information that is known about the associated image. The types of information included within such a vector may include attributes that reflect visual characteristics that are visible in an image, verbal tags that have been associated with the image users in a community of users, concepts derived from those tags, coordinates that reflect a geographic location at which a camera that produced the image was when the camera produced the image, and concepts related to groups with which the image is associated. |
| 主权项 |
1. A computer-implemented method comprising steps of:
determining, for each image in a plurality of images, a set of metadata for the image, said metadata including, for each concept of a plurality of concepts, values that indicate the probability that said each image pertains to said concept; generating, for each particular image in the plurality of images, a data structure that contains information regarding the set of metadata for the particular image; in response to a particular user's request to find other images that are similar to a selected image, comparing values in the data structure that was generated for the selected image to values in the data structure that was generated for a candidate search result image in the plurality of images; and in response to determining that a result of the comparing exceeds a specified threshold, presenting at least the candidate search result image to the user as an image that is similar to the selected image; wherein determining the set of metadata for the image comprises determining: (a) a set of attributes that reflect visual characteristics that are visible in the image, (b) a set of tags that have been associated with the image by one or more users in a community of users, (c) a set of concepts to which tags in the set of tags are related, and (d) for each concept in the set of concepts, a probability that the set of tags reflects the concept, thereby producing a set of concept probabilities for the image; wherein generating the data structure that contains information regarding the set of metadata for the particular image comprises generating a particular data structure that contains information regarding the (a) the set of attributes determined for the particular image, and (b) the set of concept probabilities determined for the particular image; for each particular tag in the set of tags determined for the particular image, (a) determining a quantity of different images, in the plurality of images, with which the particular tag is associated, and (b) weighting a value for a particular concept probability, in the data structure that was generated for the particular image, based at least part on said quantity. wherein the steps are performed by one or more computing devices. |
