LAUSR

Image segmentation is the process of dividing an image into meaningful objects to perform different analysis operations. Fuzzy connectedness (FC)-based segmentation methods usually give robust segmentation results; on the other hand, they suffer from some weaknesses. The generalized or absolute fuzzy connectivity (GFC) segmentation method is the foundation of most FC-based methods. This method has two apparent weaknesses: It combines different objects in the case of their boundaries are blurred, and it can not find the object of interest if the threshold value determined without interactive manner. In this manuscript, we introduce extensions to the GFC algorithm to tackle the mentioned weaknesses. The FC and affinity functions in the extended algorithm utilize region- and boundary-based information to overcome the first weakness. Moreover, this algorithm suggests a near optimal threshold generated automatically to eliminate the need for any interaction. Comparisons has been made to quantitatively evaluate the proposed algorithm over a three sorts of data set of scenes. Measures of relevance have been calculated for two data sets. Results indicate improved segmentation accuracy and also showed that the weaknesses of the traditional GFC algorithms have been eliminated to some extent.