Use of area-closing to improve granulometry performance

We present a new procedure for computing the granulometry function. Computation of granulometry is normally done by morphological opening/closing performed on an image with structuring elements of increasing size. This can pose problems, if pixels within the image components are not nearly uniform gray level (or have large variations) or if there are patches within the objects of interest. The proposed procedure computes the granulometry function by a two-stage process; first the sizes of patches within the image components are estimated, then they are filled and the granulometry function is computed. As an example, this method is applied to estimating the size of red blood cells in malaria affected blood slides, where pixels of the objects of interest (red blood cells) are non-flat with some patches because of the shape of the cells. The results obtained are very encouraging.