Systems and methods to present web image search results for effective image browsing

A tangible computer-readable data storage medium comprising computer-program instructions for presenting web image search results, wherein the computer-program instructions, are executed by a processor for:

• generating task-based attention objects for each of multiple images associated with image search results returned by a search engine;
• creating thumbnail images from respective ones of the images as a function of at least the task-based attention objects to emphasize image region(s) of greater priority to a user in view of a keyword or an expanded keyword associated with the search results;
• calculating similarity measurements between the images as a function of one or more of content of the task-based attention objects and features associated with the images as a whole, the features comprising one or more of features extracted from the images and context associated with the images;
• and
• mapping the thumbnail images to a 2-D layout space for presentation to the user as a function of the similarity measurements,
• wherein the similarity is measured by:
• VMij=αRMij+(1−α)IMij, wherein IM is an image similarity matrix between an ith image (Ii) and a jth image (Ij), RM is a region similarity matrix between an attention region of Ii and Ij,and α is a weight to factor in overall similarity and attention region similarity, wherein the overall similarity represents content of a mined image.

A computing device for presenting web image search results, the computing device comprising:

• a processor;
• and
• a memory coupled to the processor, the memory comprising computer-program instructions executable by the processor for:
• generating task-based attention objects for each of multiple images associated with image search results returned by a search engine;
• creating thumbnail images from respective ones of the images as a function of at least the task-based attention objects to emphasize image region(s) of greater priority to a user in view of a keyword or expanded keyword associated with the search results;
• mapping the thumbnail images to a 2-D layout space for presentation to the user,
• wherein the mapping comprises positioning thumbnail images with similar task-based attention objects in close proximity to one another and positioning thumbnail images with a higher search engine rank in front of thumbnail images with a lower search engine rank;
• calculating similarity measurements between the images as a function of one or more of content of the task-based attention objects and features associated with the images as a whole, the features comprising one or more of features extracted from the images and context associated with the images;
• and
• mapping the thumbnail images to a 2-D layout space for presentation to the user as a function of the similarity measurements,
• wherein the is similarity measured by:
• VMij=αRMij+(1−α)IMij, wherein IM is an image similarity matrix between an ith image (Ii) and a jth image (Ij), RM is a region similarity matrix between an attention region of Ii and Ij, and α is a weight to factor in overall similarity and attention region similarity, wherein the overall similarity represents content of a mined image.

A computer implemented method executed by a processor for presenting web image search results, the method comprising:

• generating task-based attention objects for each of multiple images associated with image search results from a search engine;
• creating thumbnail images from respective ones of the images as a function of at least the task-based attention objects to emphasize image region(s) of greater priority to a user in view of a keyword or an expanded keyword associated with the search results;
• calculating similarity measurements between the images as a function of one or more of content of the task-based attention objects and features associated with the images as a whole, the features comprising one or more of features extracted from the images and context associated with the images;
• and
• mapping the thumbnail images to a 2-D layout space for presentation to the user as a function of the similarity measurements,
• wherein each of the thumbnail images are fit into a grid while substantially maximizing original similarity relationships, and
• wherein the similarity is measured by:
• VMij=αRMij+(1−α)IMij, wherein IM is an image similarity matrix between an ith image (Ii) and a jth image (Ij), RM is a region similarity matrix between an attention region of Ii and Ij, and α is a weight to factor in overall similarity and attention region similarity, wherein the overall similarity represents content of a mined image.