Increasing retrieval performance of images by providing relevance feedback on word images contained in the images

A method for increasing retrieval performance of images depicting text by allowing users to provide relevance feedback on words contained in the images, the method comprising:

• receiving a query input comprising one or more query terms from a user through a user interface to locate one or more images included in a system;
• determining one or more text suggestions related to the one or more query terms, wherein the one or more text suggestions comprise terms that have been identified as being included in the one or more images;
• retrieving one or more word images from the one or more images related to the one or more text suggestions, wherein each word image is a graphical rendering of a portion of the one or more images that shows a word;
• creating one or more word image groups from the retrieved word images using one or more similarity measures;
• wherein the one or more similarity measures includes a location similarity measure, wherein two word images are similar when their bounding box positions within their different corresponding images overlap by more than a given threshold;
• determining at least one representative word image for each word image group;
• displaying the representative word image for each word image group to the user via the user interface;
• and
• receiving a selection of one or more of the displayed representative word images from the user via the user interface.

A program of instructions stored in a non-transitory computer readable medium executable by a computer to perform a function for increasing retrieval performance of images depicting text by allowing users to provide relevance feedback on words contained in the images, the function comprising the steps of:

• receiving a query input comprising one or more query terms from a user through a user interface to locate one or more images included in a system;
• determining one or more text suggestions related to the one or more query terms, wherein the one or more text suggestions comprise terms that have been identified as being included in the one or more images;
• retrieving one or more word images from the one or more images related to the one or more text suggestions, wherein each word image is a graphical rendering of a portion of the one or more
• images that shows a word;
• creating one or more word image groups from the retrieved word images using one or more similarity measures;
• wherein the one or more similarity measures includes a location similarity measure, wherein two word images are similar when their bounding box positions within their different corresponding images overlap by more than a given threshold;
• determining at least one representative word image for each word image group;
• displaying the representative word image for each word image group to the user via the user interface;
• and receiving a selection of one or more of the displayed representative word images from the user via the user interface.