Bin clustering of 3D RGB histogram

Hi and good day, Actually, I am going to do the tissue segmentation in an image taken from Diabetic foot ulcer (e.g. attached file) and I am going to implement a proposed method by Berris (1997) in which he did the tissue segmentation of different tissue types within the wound site: granulation tissue (beefy red color), slough (yellow creamy) and necrotic tissue(black)by using color histogram clustering technique as follow: 1.creating 3D RGB color histogram (that can be a cube of 256*256*256 bins or 64*64*64 bins or 16*16*16 bins and etc.) 2.smoothing the 3D RGB histogram 3.erosion and dilation of the histogram 4.creating clusters in the histogram. I've already done the first 3 steps and going to implement the last step as follow: a.find the bin with the largest count (frequency) in the histogram (already done) and looking all its connected bins (26 neighboring bins) with a given percentage of this peak value. so the first cluster will be formed. b.the subsequent clusters will be formed in similar manner,but with the added criterion that any new bins must not already be assigned to a previous cluster. In this way, each (R,G,B) Co-ordinate was designated as belonging to a specified cluster. c. once a sensible number of clusters are formed, the algorithm will scan the whole of the original image and create a segmentation color image for each cluster. so, at the end we should have number of color images,one for each cluster. So, I really appreciate if anybody has idea to implement the 4th step (a. to c.) especially the bin clustering. I am lloking forward hearing from you soon. bests