METHOD FOR PROCESSING IMAGES Russian patent published in 2008 - IPC G06T7/00 

FIELD: methods for processing visual information, possible use in computer vision systems for solving problems of searching, tracking, guidance, diagnostics and control.

SUBSTANCE: in accordance to the method, original image is divided onto sections of even areas, each one of which is matched with a matrix of "visual" masses, layers of description of image of first pyramidal level by multi-component vector of gradients on top of absolutely smooth equation-definable class are formed. Structural elements are singled out as well as connections between components of aforementioned vector with creation of first level of pyramid of descriptions, informational sub-areas of attention are selected on analyzed image, with repeated description of attention sub-area with usage of aforementioned actions for forming the next level of description hierarchy.

EFFECT: simplified image processing method due to exclusion of construction of pyramid of image copies and increased trustworthiness with set precision of produced results of image description due to introduction of system of transformations which satisfy requirements of completeness and simplicity of realization.

2 cl, 4 dwg