In my project i want to find contour subsets of an image, each delimited by two concavities. The size of every foreground region delimited by a detected contour subset and by the straight line segment connecting the two extremes of the contour subset is computed. If the size is smaller than a given threshold, the peripheral part is regarded as noisy and is removed. As for the threshold, we use the same value adopted during removal of small 8-connected foreground components and filling of thin hollows and small 4-connected holes, i.e., we remove peripheral regions with less than 32 pixels.

Please anyone can help me to find this?