Method of bringing locally varying images into coincidence in video capillaroscopy

摘要

This paper presents the results of a study of a method for estimating and compensating mutual displacements of locally varying images of a capillary network when a video sequence is being recorded in order to determine the flow rate of erythrocytes in the capillaries of a nail bed. A method is proposed for bringing images into coincidence in a video sequence, based on the use of a set of reference frames, relative to which the displacement of each image is estimated. It is shown that the proposed method reduces the (subpixel) errors in image alignment, such as local image deformations due to the nature of living biological tissue. Experimental estimates are obtained for the image-alignment error, which equalled 0.11 pixel when 30 reference frames were used. The proposed method has the advantages that it is highly noise resistant and has high computational efficiency in solving problems of video capillaroscopy. (C) 2020 Optical Society of America