Watershed Segmentation

Back to MIB | User interface | Menu | Tools

Semi-automated image segmentation using the Watershed method in Microscopy Image Browser.

Overview

The Watershed segmentation tool in MIB provides semi-automated segmentation using the Watershed method, ideal for separating objects with distinct boundaries, such as membrane-enclosed organelles.

Note

For better interactivity and performance consider using the Graphcut segmentation tool instead.

Mode panel

The Mode panel lets you select the segmentation scope.

Mode panel options

2D, current slice only segments only the current slice in the Image View panel
2D, slice-by-slice applies 2D segmentation to each slice individually
3D, volume performs 3D segmentation on the entire dataset or a subarea (see Subarea panel below)
Aspect ratio for 3D... displays the dataset’s aspect ratio, derived from voxel sizes in (Menu → Dataset → Parameters). this is used when Watershed relies on a distance map (see Image segmentation settings below)

Subarea panel

The Subarea panel defines a dataset subset for processing, useful for large datasets or binning.

Subarea panel settings

X:... sets the width range (e.g., 1:512)
Y:... sets the height range
Z:... sets the z-slice range
from Selection fills X, Y, and Z with coordinates from the Selection layer’s bounding box
Current View limits X and Y to the visible area in the Image View panel
Reset restores full dataset dimensions
Bin x times... applies a binning factor to reduce detail for faster processing

Image segmentation settings

The Watershed workflow uses labeled areas (Background and Object) for segmentation, contrasting with the Graphcut workflow. It’s less interactive, slower per execution, and best for objects with distinct boundaries. Graphcut, with its preprocessing focus, is faster for repeated interactions and handles both boundaries and intensity contrast, making it preferable for most cases.

Color channel selects the channel for segmentation
Background selects the model material labeling background areas
Object selects the model material labeling objects
Type of signal chooses black-on-white (dark boundaries) or white-on-black (bright boundaries)
Update lists refreshes material lists

Optional pre-processing (Watershed only):

Gradient applies a gradient filter to enhance object borders
Eigenvalue of Hessian preprocesses data for improved Watershed results; adjust with Sigma fields
Export to MATLAB sends preprocessed data to the MATLAB workspace
Preview displays preprocessing results in the Image View panel
Pre-process starts preprocessing (turns green when data is ready)
Import from MATLAB loads a dataset from the MATLAB workspace for segmentation
Clear removes preprocessed data from memory

Image segmentation example

Follow these steps to segment mitochondria using Watershed:

Load a sample dataset: Menu → File → Import image from → URL, enter
http://mib.helsinki.fi/tutorials/WatershedDemo/watershed_demo1.tif
Add a material named Background in the Segmentation panel with + (right-click to rename)
Use the Brush tool to label cytoplasm, then press A to add it to Background

Snapshot