Text domain image retrieval

A method, comprising:

• receiving, an image query generated by a remote device, the image query comprising image data;
• determining one or more image descriptors based on the image data, said determining comprising:
• deriving a plurality of a feature metrics for the image data;
• and
• selecting, from a descriptor repository that stores many-to-many mappings between image descriptors and text descriptors, each of the one or more image descriptors based on a comparison of the feature metrics of said image descriptor with the feature metrics of the image data;
• obtaining one or more text descriptors, said obtaining comprising, for each of the one or more determined image descriptors, obtaining the one or more text descriptors mapped to that image descriptor in the descriptor repository;
• generating an document query comprising a search parameter, the search parameter including the one or more text descriptors;
• identifying, in a document database, based on the document query, document identifiers of text documents;
• identifying, in a file mapping repository, image identifiers mapped to the document identifiers;
• and
• communicating an image query result to the remote device, the image query results indicative of images corresponding to the image identifiers.

A method, comprising:

• generating a descriptor mapping for a descriptor repository by:
• receiving training image data and training text data,
• determining an image descriptor based on the training image data, said determining comprising:
• deriving a plurality of a feature metrics for the image data;
• and
• selecting, from a descriptor repository that stores many-to-many mappings between image descriptors and text descriptors, each of the one or more image descriptors based on a comparison of the feature metrics of said image descriptor with the feature metrics of the image data,
• generating a descriptor mapping comprising an association between the image descriptor and a text descriptor, and
• storing the descriptor mapping in the descriptor repository;
• generating a text document for a document database by:
• determining the image descriptor is descriptive of a source image,
• retrieving, based on the descriptor mapping, the text descriptor associated with the image descriptor,
• compiling the text descriptor into a new text document, and
• storing the new text document in the document database;
• storing a file mapping in a file mapping repository, the file mapping comprising an identifier of the source image and an identifier of the new text document;
• responding to an image query comprising search image data by:
• selecting the image descriptor based on a comparison of features of the search image data with features of the image descriptor,
• generating a document query comprising the text descriptor associated with the image descriptor in the descriptor mapping,
• receiving, based on the document query, a document identifier of the new text document stored in the document database,
• obtaining the image identifier associated with the document identifier in the file mapping repository;
• and
• communicating an image query result to a remote device, the image query result indicative of the image corresponding to the image identifier.