doi:10.1016/j.imavis.2006.07.014 
Copyright © 2006 Elsevier B.V. All rights reserved.
Using resolution pyramids for watershed image segmentation
aInstitute of Cybernetics “E. Caianiello”, Italian National Research Council (CNR), Via Campi Flegrei 34, 80078 Pozzuoli (Naples), Italy
Received 25 November 2005;
revised 26 June 2006;
accepted 12 July 2006.
Available online 21 August 2006.
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Abstract
In this paper we build a shape preserving resolution pyramid and use it in the framework of image segmentation via watershed transformation. Our method is based on the assumption that the most significant image components perceived at high resolution will also be perceived at lower resolution. Thus, we detect the seeds for the watershed transformation at a low resolution, and use them to distinguish significant and non-significant seeds at any selected higher resolution. In this way, the watershed partition obtained at the selected pyramid level will include only the most significant components, and over-segmentation will be considerably reduced. Segmentations of the image at different scales will be available. Moreover, since the seeds can be detected at different pyramid levels, alternative segmentations of the image at a given resolution can be obtained, each characterized by a different level of detail.
Keywords: Segmentation; Watershed transformation; Resolution pyramid
Fig. 1. The 256 × 256 grey-level image used as running example.
Fig. 2. The eight neighbours of p, top, and the four 2 × 2 blocks that can be obtained by shifting the sampling grid.
Fig. 3. Multiplicative masks of weights used to build the pyramid.
Fig. 4. The four levels of the pyramid computed for the running example.
Fig. 5. The watershed lines found at the four pyramid levels starting from the relative regional minima are superimposed on the images at all pyramid levels.
Fig. 6. The watershed lines found at the four pyramid levels after region merging.
Fig. 7. From left to right: seeds projected from levels 32 × 32, 64 × 64, and 128 × 128, onto level 256 × 256.
Fig. 8. Left, projected seeds are black, original seeds non overlapping the projected seeds are grey, and original seeds overlapping the projected seeds are white. Right, magnified top-left portion of the image.
Fig. 9. Seeds found at level 256 × 256, left, and significant seeds remaining after seed removal based on Criterion I.
Fig. 10. Watershed partitions at level 256 × 256, obtained by using the seeds projected from levels 128 × 128, top-left, 64 × 64, top-right, and 32 × 32, bottom.
Fig. 11. Shrunk projected seeds at level 256 × 256, left, and significant seeds remaining at level 256 × 256, after the seed removal process based on Criterion II, right.
Fig. 12. Watershed partition at level 256 × 256 obtained by using the seeds projected from levels 128 × 128, top-left, 64 × 64, top-right, and 32 × 32, bottom, respectively.
Fig. 13. Partition obtained at the segmentation level 128 × 128, by using the seeds projected from level 64 × 64 and Criterion II
Fig. 14. Segmentation at level 256 × 256, obtained by projecting the basic watershed transform computed at level 64 × 64. Note the rather squared edges.
Fig. 15. For each row from left to right: input image at level 256 × 256, and segmentations obtained by using 128 × 128, 64 × 64 and 32 × 32, respectively, as seed level. Criterion II is used for seed removal and region merging involves only adjacent regions whose representative grey-levels differ at most by 10.
Abstract
Purchase PDF (2666 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (3)
Purchase PDF (645 K)
