Apparatus and method for image recognition

Abstract

An apparatus, method and computer program for recognizing one or more images within digitized image data that might include a target image desired to be located. The system generates a set of domain blocks from the image data where each domain block represents a discrete portion the image data and a set of range blocks from one or more target images. Either the domain blocks, the range blocks, or both, are transformed by one or more substantially affine transformations with predetermined coefficients to create possible variants of the images. A comparison between the blocks is made to determine similarity, and includes at least a measurement of whether better matching is achieved when a range block is chosen from image data representing the image which is the source of the domain block or when chosen from other image data.

1 - 32 . canceled  
   
   
       33 . An apparatus for unconstrained recognition of one or more images within digitized image data, comprising at least one processor that: 
 receives image data in a digital format;    generates a set of domain blocks from the image data, each domain block representing a discrete portion the image data;    generates a set of range blocks from a predetermined one or more target images that are desired located within the image data, the range blocks corresponding to discrete portions of the one or more target images, and the range blocks are transformed by one or more substantially affine transformations with predetermined coefficients;    compares each domain block with one or more of the range blocks;    while comparing, generates classification data based upon a measurement of whether the comparison is closer when one or more range blocks represent at least a portion of the one or more target images, or the comparison is closer when one or more range blocks represent at least a portion of the image data representing the image which is the source of the domain block, the classification data including an adjustment for error;    determines a similarity indication of at least a specific portion of one or more target images being similar to specific portions of the image data based upon the classification data; and    provides the similarity indication to a third party application.    
   
   
       34 . The apparatus of  claim 33 , wherein the third party application is a database.  
   
   
       35 . The apparatus of  claim 33 , wherein the third party application is a repository accessible through a network.  
   
   
       36 . The apparatus of  claim 35 , wherein the network is the Internet.  
   
   
       37 . The apparatus of  claim 33 , wherein the third party application is a search engine.  
   
   
       38 . The apparatus of  claim 33 , wherein the target images are images of known terrorists.  
   
   
       39 . The apparatus of  claim 33 , wherein the target images are images of known criminals.  
   
   
       40 . The apparatus of  claim 33 , wherein the target images are faces of known individuals.  
   
   
       41 . The apparatus of  claim 33 , wherein the target images are images of transportation vehicles.  
   
   
       42 . The apparatus of  claim 41 , wherein the transportation vehicles are aircrafts.  
   
   
       43 . The apparatus of  claim 41 , wherein the transportation vehicles are ships.  
   
   
       44 . The apparatus of  claim 33 , wherein the target images are satellite images.  
   
   
       45 . The apparatus of  claim 33 , wherein the target images are reconnaissance images.  
   
   
       46 . A method for unconstrained recognition of one or more images within digitized image data, comprising the steps of: 
 digitizing image data;    generating a set of domain blocks from the image data, each domain block representing a discrete portion the image data;    generating a set of range blocks from a predetermined one or more target images that are desired located within the image data, the range blocks corresponding to discrete portions of the one or more target images, and the range blocks are transformed by one or more substantially affine transformations with predetermined coefficients;    comparing each domain block with one or more of the range blocks;    while comparing, generating classification data based upon a measurement of whether matching is achieved when a range block representing at least a portion of the one or more target images is similar to a domain block, and at least a measurement of whether better matching is achieved when a range block is chosen from image data representing the image which is the source of the domain block, the classification data including an adjustment for error;    determining a similarity indication of at least a specific portion of one or more target images being similar to specific portions of the image data based upon the classification data; and    providing the similarity indication to a third party application.    
   
   
       47 . The apparatus of  claim 46 , wherein the third party application is a database.  
   
   
       48 . The apparatus of  claim 46 , wherein the third party application is a repository accessible through a network.  
   
   
       49 . The apparatus of  claim 48 , wherein the network is the Internet.  
   
   
       50 . The apparatus of  claim 46 , wherein the third party application is a search engine.  
   
   
       51 . The apparatus of  claim 46 , wherein the target images are images of known terrorists.  
   
   
       52 . The apparatus of  claim 46 , wherein the target images are images of known criminals.