A Concept Of Evaluation Of Shape And Cell's Texture By Space Vectors Positions
A. Houri, L. Fongang, A. Bijaoui · 2005
Cells caracterization by mathematical morphology analy- sis of comsponding shape and biological texture images is describe. We suggest in this paper, a method based on cell's spatial reorientation by vectors stemming from his centre of gravity and rests on his contour. This method point out an simple way to get similar positions between two (or more) cells (or two domains of the same cell), during their comparizon. Analysis and comparizon parameters are derive from second-order statistics for stochastics textures analysis and, from chi-square distribution to capture struc- tures when co-occurrence matrix is used over structured (deter- niinistically or randomly) local subpattems. The method applicated to blood cells, gives an high iden- tification accuracy ; some examples are presented. MTRODUCTION. Textural and morphologycal cells caracterization of cor- responding digital images, by mathmatical analysis, requiered seve ral methods. For example : - cross-correlation - mathematical morphology Those algorithms are not sufficient to define an high rec- ognition accuracy, with comparizon to biologists visual analysis. In order to acheave morphologycal identification, the classification obtained using both methods must be improved. To fit mathemathi- cal identification to the biologists visual analysis, we added several evaluations tests. The laters consider textures, shapes and cells orientations scattering, relative to the intensity variations that oc- cured either in exposure time or during image digitalization, and then increase identification difficulties.