Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating and providing animation snippets. In one aspect, a method includes determining page resources that each include a plurality of persistent visual resources, for each of the determined page resources, generating one or more animation snippets from the persistent visual resources, for a page resource for which one or more animation snippets have been generated, selecting one of the animation snippets for display as a search result that identifies the page resource as being responsive to a search query; and providing animation snippet to a user device from which the query was submitted.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating descriptive text for images. In one aspect, a method includes identifying a set of seed descriptors for an image in a document that is hosted on a website. For each seed descriptor, structure information is generated that specifies a structure of the document with respect to the image and the seed descriptor. One or more templates are generated for each seed descriptor using the structure information for the seed descriptor. Each template can include image location information, document structure information, image feature information, and a generative rule that generates descriptive text for other images in other documents. Descriptive text for other images is generated using the templates and the other documents. The descriptive text is associated with the images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating descriptive text for images. In one aspect, a method includes identifying a set of seed descriptors for an image in a document that is hosted on a website. For each seed descriptor, structure information is generated that specifies a structure of the document with respect to the image and the seed descriptor. One or more templates are generated for each seed descriptor using the structure information for the seed descriptor. Each template can include image location information, document structure information, image feature information, and a generative rule that generates descriptive text for other images in other documents. Descriptive text for other images is generated using the templates and the other documents. The descriptive text is associated with the images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating and providing animation snippets. In one aspect, a method includes determining page resources that each include a plurality of persistent visual resources, for each of the determined page resources, generating one or more animation snippets from the persistent visual resources, for a page resource for which one or more animation snippets have been generated, selecting one of the animation snippets for display as a search result that identifies the page resource as being responsive to a search query; and providing animation snippet to a user device from which the query was submitted.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating descriptive text for images. In one aspect, a method includes identifying a set of seed descriptors for an image in a document that is hosted on a website. For each seed descriptor, structure information is generated that specifies a structure of the document with respect to the image and the seed descriptor. One or more templates are generated for each seed descriptor using the structure information for the seed descriptor. Each template can include image location information, document structure information, image feature information, and a generative rule that generates descriptive text for other images in other documents. Descriptive text for other images is generated using the templates and the other documents. The descriptive text is associated with the images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for adjusting facial image search results. In one aspect, a method includes detecting, in images that are responsive to a query, faces in the images and building facial templates. The images are then clustered according to similarity measures of the facial templates. The cluster with the highest quantity is selected, and each image in the set of returned images is compared to each image in the selected cluster. The similarity of each face with the faces in the largest cluster, based on the facial templates, is determined as an inlier score for the image in which the face is depicted. The system then increases the rank of images with high inlier scores and decreases the rank of images with low inlier scores.

Abstract: Methods, systems, and apparatus for identifying landing pages for images. In one aspect, a method includes, for each image of a plurality of images, identifying web pages that each include the image, and, for each of the web pages, receiving a web score generated by a ranking algorithm. The web page having a web score indicating a highest importance relative to other web pages is selected, sets of similar images are determined from the images included in the selected web pages, and each set of similar images includes different images. For each set of similar images, an image is selected based on the selected web page that includes the selected image having a highest web score of the set, and the selected web page that includes the selected image is identified as a landing page for an image search result.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for adjusting facial image search results. In one aspect, a method includes detecting, in images that are responsive to a query, faces in the images and building facial templates. The images are then clustered according to similarity measures of the facial templates. The cluster with the highest quantity is selected, and each image in the set of returned images is compared to each image in the selected cluster. The similarity of each face with the faces in the largest cluster, based on the facial templates, is determined as an inlier score for the image in which the face is depicted. The system then increases the rank of images with high inlier scores and decreases the rank of images with low inlier scores.

Abstract: Methods, systems, and apparatus for identifying landing pages for images. In one aspect, a method includes, for each image of a plurality of images, identifying web pages that each include the image, and, for each of the web pages, receiving a web score generated by a ranking algorithm. The web page having a web score indicating a highest importance relative to other web pages is selected, sets of similar images are determined from the images included in the selected web pages, and each set of similar images includes different images. For each set of similar images, an image is selected based on the selected web page that includes the selected image having a highest web score of the set, and the selected web page that includes the selected image is identified as a landing page for an image search result.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating descriptive text for images. In one aspect, a method includes identifying a set of seed descriptors for an image in a document that is hosted on a website. For each seed descriptor, structure information is generated that specifies a structure of the document with respect to the image and the seed descriptor. One or more templates are generated for each seed descriptor using the structure information for the seed descriptor. Each template can include image location information, document structure information, image feature information, and a generative rule that generates descriptive text for other images in other documents. Descriptive text for other images is generated using the templates and the other documents. The descriptive text is associated with the images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for identifying click magnet images. In one aspect, a method includes obtaining selections vector for images. A selection vector for an image can include vector elements that correspond to a unique search query. The value of each vector element can be proportional to a number of selections of image search results that included a representation of the image when the search results were presented in response to the unique search query. The image can be deemed be a click magnet image based at least in part on a first number of selections of image search results that included a representation of the image for search queries categorized as belonging to a set of categories and a total number of selections of image search results that included a representation of the image.

Abstract: A system for performing visible object determination. Visualization software executing on one or more processors searches a cone tree with respect to a hull tree. Both trees respect the same group of transformations. Thus, the cone tree as stored in memory need not store all branches down to ultimate resolution. For example, the cone tree may store a selected cone and all its subcones down to the ultimate resolution, while subcones of other cones at the same level as the selected cone may not be represented in the stored cone tree. When the visualization software encounters a nonstored cone and a given hull H in its search procedure, the visualization software may apply an appropriate one of the group transformations to the given hull, and search the tranformed hull with respect to the selected cone (or a subcone of the selected cone).

Abstract: A system for performing visible object determination. Visualization software running on a host processor represents space with a hierarchy of cones, and generates a hierarchy of bounding hulls from a collection of objects by recursively grouping clusters of objects. Each hull node in the hull hierarchy stores parameters which characterize a bounding hull for the corresponding cluster or object. The visualization software searches the cone and hull hierarchies starting with the root cone and the root hull. Before exploring a given cone-hull pair, a normalized cone size for the cone and a normalized hull size for the hull may be computed, and the sizes compared. If the cone size is larger than the hull size, subcones of the cone are explored with respect to the hull. Conversely, if the hull size is larger than the cone size, subhulls of the hull are explored with respect to the cone.

Abstract: A system and method for determining graphical object visibility by utilizing ancillary geometry. When comparing visibility cones with bounding hulls of scene graph hierarchy leaf nodes, ancillary geometry may be used in place of the bounding hulls and the leaf node's actual geometry. The ancillary geometry may be created through a manual process, or by an automated decimation process. The ancillary geometry may be a simplistic analytic surface (e.g., a sphere or a cylinder), a NURB surface, a collection of polygons, or a combination of surfaces and polygons.

Abstract: A system for performing visible object determination. Visualization software executing on one or more processors searches a cone tree with respect to a hull tree. Both trees respect the same group of transformations. Thus, the cone tree as stored in memory need not store all branches down to ultimate resolution. For example, the cone tree may store a selected cone and all its subcones down to the ultimate resolution, while subcones of other cones at the same level as the selected cone may not be represented in the stored cone tree. When the visualization software encounters a nonstored cone and a given hull H in its search procedure, the visualization software may apply an appropriate one of the group transformations to the given hull, and search the tranformed hull with respect to the selected cone (or a subcone of the selected cone).

Abstract: A system for performing visible object determination. Visualization software running on a host processor represents space with a hierarchy of cones, and generates a hierarchy of bounding hulls from a collection of objects by recursively grouping clusters of objects. Each hull node in the hull hierarchy stores parameters which characterize a bounding hull for the corresponding cluster or object. The visualization software searches the cone and hull hierarchies starting with the root cone and the root hull. Before exploring a given cone-hull pair, a normalized cone size for the cone and a normalized hull size for the hull may be computed, and the sizes compared. If the cone size is larger than the hull size, subcones of the cone are explored with respect to the hull. Conversely, if the hull size is larger than the cone size, subhulls of the hull are explored with respect to the cone.

Abstract: A system and method for determining graphical object visibility by utilizing ancillary geometry. When comparing visibility cones with bounding hulls of scene graph hierarchy leaf nodes, ancillary geometry may be used in place of the bounding hulls and the leaf node's actual geometry. The ancillary geometry may be created through a manual process, or by an automated decimation process. The ancillary geometry may be a simplistic analytic surface (e.g., a sphere or a cylinder), a NURB surface, a collection of polygons, or a combination of surfaces and polygons.

Abstract: A system and method for performing visible object determination based upon a dual search of a cone hierarchy and a bound hierarchy performed by multiple processors. Each processor is configured to read a (global and/or local) problem queue to access a bound-cone pair. The bound-cone pair points to a bound in the bound hierarchy and a cone in the cone hierarchy. The processor computes a bound-cone distance between the bound and the cone, and compares the bound-cone distance to a visibility distance associated with the cone. If the bound-cone distance is smaller than the visibility distance, the processor may write two or more refined bound-cone pairs corresponding to a refinement of the original pair to the global or local problem queue. When the processor detects a leaf bound and a leaf cone, it updates a nearest object pointer and the visibility distance associated with the leaf cone.

Abstract: A system for determining nearest leaf bounds per leaf cone of a cone hierarchy based on a dual-tree search of the cone hierarchy and a bound hierarchy. A processor, performing the dual-tree search, maintains for each leaf cone a collection of nearest leaf bounds. The collection evolves as the dual-tree search progresses. The processor may maintain the N closest leaf bounds for each leaf cone. The processor may classify leaf bounds as occluders or non-occluders, and store any non-occluder closer than the Nth occluder. Alternatively, the processor may assign an occlusion metric value to each leaf bound, and may maintain closest leaf bounds so that the sum of their occlusion metric values is less than or equal to an occlusion threshold. The dual-tree search may be followed by a search of the bound hierarchy with probe cones of the leaf cones.

Abstract: A system for performing visible object determination. Visualization software running on a host processor represents space with a hierarchy of cones, and generates a hierarchy of bounding hulls from a collection of objects by recursively grouping clusters of objects. Each hull node in the hull hierarchy stores parameters which characterize a bounding hull for the corresponding cluster or object. The visualization software searches the cone and hull hierarchies starting with the root cone and the root hull. Before exploring a given cone-hull pair, a normalized cone size for the cone and a normalized hull size for the hull may be computed, and the sizes compared. If the cone size is larger than the hull size, subcones of the cone are explored with respect to the hull. Conversely, if the hull size is larger than the cone size, subhulls of the hull are explored with respect to the cone.