A Verification-Based Multithreshold Probing Approach to HEp-2 Cell Segmentation.

In this paper we propose a novel approach to HEp-2 cell segmentation based on the framework of verification-based multithreshold probing. Cell hypotheses are generated by binarization using hypothetic thresholds and accepted/rejected by a verification procedure. The proposed method has the nice property of combining both adaptive local thresholding and involvement of high-level knowledge. We have realized a prototype implementation using a simple rule-based verification procedure. Experimental evaluation has been performed on two public databases. It is shown that our approach outperforms a number of existing methods.