FEATURE EXTRACTION METHOD AND APPARATUS Russian patent published in 2018 - IPC G06K9/46 

FIELD: image processing means.

SUBSTANCE: invention relates to image processing. Image extraction method includes: dividing the image into a plurality of blocks, each of said blocks comprising a plurality of cells; sparse signal decomposition on each cell by using a predetermined dictionary so as to obtain a sparse vector for each cell, wherein the predetermined dictionary D is a dictionary determined by applying an iterative algorithm to a sample image. Extracting a histogram of oriented gradients of the image according to the sparse vector, wherein extraction of of the histogram of oriented gradients of the image according to sparse vectors includes: calculating the magnitude and direction of the gradient of each cell in accordance with sparse vectors to obtain the descriptor of each cell; obtaining statistics on the corresponding descriptors in each block to obtain a histogram of oriented gradients of each block; obtaining statistics on the histogram of oriented gradients of each block to obtain a histogram of oriented gradients of the image.

EFFECT: technical result is extraction of the histogram of oriented gradients.

13 cl, 12 dwg