METHOD FOR RECOGNITION OF CELL TEXTURE IMAGE Russian patent published in 2010 - IPC G06T7/40 

FIELD: medicine.

SUBSTANCE: proposed method of analysis of object texture image for detection of pathology includes identification of object texture criteria, which consists in the fact that received images of object are pre-processed by means of segmentation on the basis of analysis in colour space of object image areas, which are informative from the point of view of texture criteria and detected by selected size of window, pitch and direction of object viewing direction, values are calculated for texture criteria and numbers matrices are formed, and object is identified and classified, at that texture criteria are detected for images of blood cells, pre-processing of obtained images of object by means of segmentation is carried out on the basis of cluster analysis in colour space, areas of object image that are informative from the point of view of texture criteria are represented by selection of blood nuclei images, criteria, which is most informative for classification are detected automatically on the basis of training selections from previously generated bases of images, classified by experts and containing cells of each type (blast cells, lymphocytes), on the basis of which identification and classification are carried out, image pixels are transformed into various colour spaces, matrix of spatial adjacency is calculated, afterwards, on the basis of texture criteria calculated values, classification is carried out (distribution according to common criteria) in space of texture criteria according to the following indices: energy, inertia torque, maximum probability of adjacencies, which are met most frequently in this image, local homogeneity, entropy, trace of normalised matrix of spatial adjacency (NMSA), average brightness value, correlation of image brightness values.

EFFECT: method improvement.

9 cl, 3 dwg