Object identification apparatus and object identification method

An apparatus for classifying an input image, comprising:

• at least a processor and memory coupled to each other and cooperating to serve as:
• a recording unit configured to record a plurality of registered images which are classified into a plurality of categories based on attributes of objects included in respective ones of the registered images;
• a first extraction unit configured to extract, from standard partial feature regions in objects, feature amounts required to classify a plurality of images including the objects into the plurality of categories, wherein the standard partial feature regions are commonly defined for all of the plurality of categories;
• a second extraction unit configured to extract feature amounts, from adaptive partial feature regions in objects in images, wherein the adaptive partial feature regions are defined for each pair of any two categories among the plurality of categories, respectively, and the adaptive partial feature regions corresponding to a pair of two specified categories among the plurality of categories are used to extract the feature amounts required to classify an image belonging to any one of the two specified categories into one of the two specified categories;
• a correlation value calculation unit configured to respectively calculate first correlation values which represent correlations between feature amounts extracted from the standard partial feature regions in the input image, and feature amounts extracted from the standard partial feature regions in a registered image to be processed using said first extraction unit, and second correlation values which represent correlations between feature amounts extracted from the adaptive partial feature regions in the input image and feature amounts extracted from the adaptive partial feature regions in the registered image to be processed using said second extraction unit;
• a similarity calculation unit configured to calculate a plurality of similarities between the registered image to be processed and the input image based on the first correlation values and the second correlation values calculated by said correlation value calculation unit in correspondence with respective categories of the plurality of categories except for the category to which the registered image to be processed is classified;
• and
• an output unit configured to output an integrated similarity between the input image and the registered image to be processed by integrating the plurality of similarities calculated in correspondence with respective categories except for the category to which the registered image to be processed is classified.