FIELD: data processing.
SUBSTANCE: present invention relates to image segmentation. In one embodiment, a computer-implemented three-dimensional image segmentation method is provided. Method may include receiving a three-dimensional image obtained by an image forming apparatus and selecting a plurality of stacks of adjacent two-dimensional images from the three-dimensional image. Method may further include segmentation by a processor of each stack of adjacent two-dimensional images using a neural network model. Method may also involve determining, by a card processor, marks for a three-dimensional image by aggregating segmentation results from a plurality of stacks.
EFFECT: technical result consists in improvement of accuracy of automatic segmentation of images.
24 cl, 8 dwg
