FIELD: image processing means.
SUBSTANCE: invention relates to computing image descriptors. Method for converting an image descriptor into a converted image descriptor comprises the steps of: receiving an image descriptor basing on a histogram (h) of gradients, wherein the histogram of the gradients contains a plurality of the histogram components (hi) containing the first set of values, which are calculated in sample points of the image in the area near the key point of the image and are summarized into orientation histograms that summarize the content as per subareas of the image; applying the location criterion of the said histogram components to convert the said gradient histogram, herewith the said location criterion is provided for calculating the second set of values representing the sample image points near the said key image point extracted from the gradient histogram components, which demonstrate a variety of angular distances; generating a converted image descriptor containing the said second set of values.
EFFECT: reducing the size of the image descriptor by conversion.
17 cl, 20 dwg
