| 摘要 |
<p>Provided is an image retrieval device capable of accurately retrieving an image to be retrieved that a user intends from an input image in a short time. An image retrieval device (100) compares the feature amount of an image to be retrieved based on the attribute of the image to be retrieved which is subject to retrieval with the feature amount of an input image based on the attribute of the input image to extract an image approximate to the input image. In this case, each of the feature amounts consists of a first feature amount representing a global relative positional relationship of each pixel constituting the image to the pixels of the entire image and a second feature amount representing a local relative positional relationship to the pixels adjacent to each pixel. Furthermore, the image retrieval device (100) stores constitutive images constituting a handwritten input image inputted by the user in the past and combination images generated by combining the constitutive images in association with the image to be retrieved, compares the feature amount in each of these images with the feature amount of each stroke of the user, and displays an image to be retrieved corresponding to a constitutive image or combination image with the distance between the images short.</p> |
