METHOD FOR TWO-DIMENSIONAL HIERARCHICAL DISCRETE FILTERING OF OBJECTS Russian patent published in 2025 - IPC G06V10/25 

FIELD: image processing.

SUBSTANCE: invention relates to image processing and can be used in computer vision systems when solving image preprocessing tasks for finding objects thereon. Method includes generating a video signal of an image, measuring samples of the autocorrelation function and energy of the image, selecting a reference, forming a reference window with dimensions greater than the dimensions of the reference and the detected object, measuring samples of the autocorrelation function of the window of the reference, normalizing the samples of the video signal of the image and the reference window relative to the measured energy of the video signal of the image, setting the filter aperture, synthesis of the two-dimensional discrete filter, detecting objects with the synthesized filter; for each detected object, selecting a field of a given size within the selected field, re-measuring readings of the autocorrelation function and image energy, through which a new two-dimensional discrete filter is synthesized and presence of the object is specified.

EFFECT: low probability of false detection of objects due to hierarchical image processing in order to clarify presence of an object.

1 cl, 2 dwg