E.L. Maryaskin, A.P. Nemirko

Echocardiography is one of the main methods of heart visualization. But manifested dependence on researcher-s qualification is a significant drawback. The existing methods do not imply the usage of automated tools. The problems of processing image sequences as two-dimensional signals has obtained great development, this connected with the fast development of hardware. The dynamic segmentation method - distinction of moving objects on the image sequence - lays the foundation of the suggested processing method. Motion and altering of brightness level are not equivalent. Motion field is the actual movement of object on the scene, projected on the image plane. Optical flow is defined as the flow of brightness levels on the image plane. Optical flow and motion field are equal if objects do not alter energy brightness during the motion process. A number of algorithms are elaborated and realized for the calculation of optical flow. One of the best known is Lucas Kanade algorithm. Several additional constraints have to be imposed for solving such equation. The method of optical flow processing based on researches in the field of dynamic scenes segmentation is suggested as applied to echocardiography. The presented method differ from those adopted in image processing as being notable firstly for taking both vector components into consideration and secondly for working with flow field as new three-dimensional signal. The treatment process consists in repeating data transclastering, using the set of filters, which make it possible to define the real cluster scopes more exactly, on every step. Iteration repeats until required characteristics are reached. Model scenes imitating movement and interaction of objects were used in the course of the research and for corroboration of devised methods correctness. The subject of main concern in the conduct of the algorithm, results and quality of its operation for real data of heart ultrasonic diagnostics. The real recording was used. This method showed its adequacy for solving problems of echocardioscopy data dynamic segmentation. In the result we have an image, showing areas mutually moving, which is obvious enough for undersanding. This method can be proposed for using in different areas of biomedical researches.