Segmentation based on measures of contrast, homogeneity, and region size

Sankar K. Pal, Nikhil R. Pal
<span title="">1987</span> <i title="Institute of Electrical and Electronics Engineers (IEEE)"> <a target="_blank" rel="noopener" href="" style="color: black;">IEEE Transactions on Systems, Man and Cybernetics</a> </i> &nbsp;
From these two examples, we see that our P-P granule characterization of the transition matrix allows us to make inferences with knowledge structures about V other than simply the pure probability distribution. One other further benefit of this approach is that it allows us to integrate easily the multiple relations between variables, one of which may be a transition matrix. Assume V and U are two variables. Assume we have two pieces of information indicating the relationship between these
more &raquo; ... bles. The first is transition matrix T of the type we have just discussed. The second is an implication statement of the form where A and B are fuzzy subsets of X and Y. We can represent the first relationship, the transition matrix as di P-P granule can also be represented as a P-P granule of the form (V,U)ism 2 where m 2 has one focal element / / where H(x,y)=lA(l-A(x)+B(y)) and of course m 2 {H) =1. These two P relations can be conjuncted to give an overall relationship (VM)ism where m = »Ι 1 Π/Μ 2 . In this case if G k , k = 1, · · · are the focal elements of m i then the focal elements of m are H k , k =1 * · ·, where H k = G k n H and m(H k ) = m l (G k ). We could then use this new effective relationship with any information we have about V to make inferences about V. CONCLUSION We have shown how to represent transitional matrices in the framework of P-P granules. Abstract -Two algorithms are described for automatic image segmenta tion using a "homogeneity" measure and "Contrast" measure defined on the cooccurrence matrix of the image. The measure of contrast involves the concept of logarithmic response (adaptibility with background inten sity) of the human visual system. Provisions are also kept in two different ways to remove the undesirable thresholds. The effectiveness of the algorithms is demonstrated for a set of images having different types of histograms. The performance of the algorithms is compared to existing ones.
<span class="external-identifiers"> <a target="_blank" rel="external noopener noreferrer" href="">doi:10.1109/tsmc.1987.6499294</a> <a target="_blank" rel="external noopener" href="">fatcat:q3dmvnhgrvakzlnyqqbe6r2bhu</a> </span>
<a target="_blank" rel="noopener" href="" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener noreferrer" href=""> <button class="ui left aligned compact blue labeled icon button serp-button"> <i class="external alternate icon"></i> </button> </a>