Description of shape feature for cell image based curvature of discrete points
Focusing on description of shape feature for cell image, an approach describing shape feature of cell image based on curvature of discrete points and a computing method of curvature of discrete points based osculating circle radius under k-neighborhood were proposed. Firstly, initial contour of cell image was obtained by Canny edge detection algorithm, and clearances in the initial contour were mended using linear interpolation to gain a continuous, closed contour with unit width. Secondly, coordinate of osculating circle centre and radius of osculating circle were used to calculate curvature on every contour point, according to the calculable relation between curvature and radius of osculating circle. Finally, shape feature of cell image was described by the curvature of discrete points obtained above. Simultaneously, different curvatures of discrete points of a same cell image to describe its shape feature were computed under different k-neighborhood, to get an appropriate k-neighborhood where the description method could reflect more information of shape feature using least feature points after analysis and comparison. The proposed method can describe shape feature of cell image accurately, reliably and efficiently indicated by experiment.
NETL
NSTL
万方数据
