Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
532282 | Pattern Recognition | 2013 | 11 Pages |
A novel nonparametric concavity point analysis-based method for splitting clumps of convex objects in binary images is presented. The method is based on finding concavity point-pairs by using a variable-size rectangular window. The concavity point-pairs can be either connected with a straight split line or with a line that follows a path of minimum or maximum intensity on an accompanying grayscale image. Using straight lines can result in non-smooth contours. Therefore, post-processing steps that remove acute angles between split lines are proposed. Results obtained with images that have clumps of biological cells show that the method gives accurate results.
► Clump splitting method incorporating image intensity and holes present in clumps. ► Nonparametric method due to the use of rectangular window for concavity point-pair search. ► Post-processing ensures objects conforming to the ones present in the actual image. ► Accurately splits clumps of varying object sizes, probability and amount of overlap. ► Useful in wide range of applications with images containing clumps of convex objects.