Segmentation of cells with inclusions in plant cross sectional images

Computerized analysis of microscopic cross sectional images of plant parts is useful for authentication and quality assessment of plant materials. The analysis involves segmentation of cells in the cross section, extraction of cell features and classification of cells. The segmentation of individual cells plays a major role in realistic estimation of cell parameters for identification and classification of cells. However, the segmentation of the cells is adversely influenced due to presence of cell inclusions like starch grains which are stained with different colored dyes for improved visibility. During segmentation, the stained inclusions touching the boundary of the cells result in boundary cavities which deform the shape of the segmented cells. The present paper describes an image processing technique for accurate segmentation of cells in the presence of stained cell inclusions. The proposed method corrects the shape of individual cells by identifying and selective filling of the boundary cavities resulting from stained inclusions with the help of convex image of the cells and morphological skeleton of the background region. The proposed method is demonstrated by testing a plant cross sectional image containing starch grains.