Image similarity determining device and method, and an image feature acquiring device and method

An image feature acquiring device, comprising:

• a preprocessing unit configured for preprocessing an input image so as to divide the input image into at least one input image region and extract feature points of each input image region;
• a feature region constructing unit configured for constructing a feature region with respect to each feature point in a feature point set of each input image region having at least three feature points, wherein the feature region satisfies with the following conditions:
• the feature point is used as a vertex of the feature region;
• a ray formed by the vertex and at least one of the other feature points in the feature point set is used as an edge of the feature region, the feature region constructed by two of said edges includes all feature points of the feature point set, and the angle of the feature region is the smallest;
• and
• an image feature acquiring unit configured for determining geometric features of the input image region, according to the distribution of respective feature points in each of the feature regions, as a type of image feature of the input image region.

An image similarity determining method for determining similarity between an input image and a data source image, the image similarity determining method comprising:

• dividing the input image into at least one input image region and dividing the data source image into at least one image region to be matched, and extracting feature points of each input image region and each image region to be matched;
• determining, with respect to each input image region, one to one matched feature point pairs between the input image region and each image region to be matched, according to the feature similarities between each feature point of the input image region and each feature point of respective image regions to be matched, and forming a matched feature point set including all the one to one matched feature points for the input image region and its corresponding image regions to be matched according to the one to one matched feature point pairs;
• determining, with respect to each input image region and its corresponding image region to be matched, a geometry similarity between the input image region and the corresponding image region to be matched, based on the distribution of respective feature points in the matched feature point set of the input image region and the distribution of respective feature points in the matched feature point set of the corresponding image region to be matched;
• and
• determining the image similarity between each input image region and its corresponding image region to be matched according to the geometry similarity between the input image region and the corresponding image region to be matched, and determining the similarity between the input image and the data source image according to the image similarities between each of the input image regions of the input region and respective corresponding image regions to be matched.