Watershed transformation

from Wikipedia, the free encyclopedia

The watershed transformation ( WST ; Engl. Watershed transformation ) is a method for segmentation within the image processing .

principle

The method is applied to gray-scale images. The gray value is interpreted as height information. With the successive flooding of the Grauwertgebirge, watersheds are built between adjacent reservoirs. As a rule, this results in an over-segmentation of the image, in particular in the case of noisy image material, for example medical CT data. The image must either be pre-processed or the regions must then be combined in a merge step using a similarity criterion. Alternatively, depending on the application, a variant of the WST can be used.

The required gray value image is obtained, for example, by calculating the gradient of the original image; the watersheds are said to be built here later along strong edges. In the case of binary data, the inverse of the Euclidean distance transformation can be calculated. The WST should then, for example, separate related objects.

variants

In the pre-flooded WST , only reservoirs that exceed a certain size are flooded.

In the case of hierarchical WST , the result is converted into a graph representation (that is, the neighborhood relationships of the segmented regions are determined) and further WST are then carried out recursively. Problem: The watersheds are getting wider and wider.

With the marker-based WST , the flooding takes place only from certain marker positions that the user has previously set interactively or that have been obtained beforehand using morphological operators.

With interactive variants of the WST it is possible to set so-called include and exclude points or to set up artificial watersheds in order to improve the segmentation result.

software

The algorithm is implemented in the free image processing libraries Scikit-image and OpenCV .

Individual evidence

  1. ^ Joachim Ohser, Katja Schladitz: 3D Images of Materials Structures: Processing and Analysis . Wiley-VCH Verlag, ISBN 352731203X
  2. ↑ Markers for watershed transform - skimage v0.15.dev0 docs. Retrieved September 13, 2018 .
  3. ^ OpenCV: Image Segmentation with Watershed Algorithm. Retrieved September 13, 2018 .