Loading...
 

Thresholding

Principles

Thresholding is a common technique to segment object by using only pixel values. A threshold is defined, all pixels with values higher (in the case of bright spots) that the threshold are considered to be a part of an object, pixels with values lower than this threshold are considered to be part of background. The threshold can be set up manually or using many algorithms that try to determine the best threshold for each images.

It can be important to reduce noise prior applying thresholding.
When images are very noisy, a good strategy for spot detection is to apply a threshold on the hessian transform (inverted)

Simple Segmenter

This plugin is a simple thresholding plugin, you can specify a low threshold using different algorithms. You can use a constant value, or a autoThreshold as implemented in ImageJ (see this link). You can also specify a high threshold, or use by default the max threshold value.
You can also a specify a size range using minimum size and eventually specify a maximum size or use the default max size.

Hysteresis Segmenter

This technique actually define 3 classes using two thresholds. Pixels below low threshold are considered as background, pixels above high threshold are considered as objects and pixels between low and high thresholds are considered as intermediate and will be linked to background or objects depending on their neighbourhood. All intermediate pixels connected to at least one pixel above high threshold are transformed into object pixels, otherwise they are transformed to background pixels.
This plugin uses the hysteresis threshold algorithm implemented in imagescience.

Hysteresis schema
Hysteresis schema

Blue areas are intermediate pixels (between low and high thresholds), red areas are objects pixels (higher than high threshold). Upper row, thresholded areas, lower row, results of hysteresis segmentation, blue area touching red area is considered as object, blue area on the left is not connected to red pixels and will be removed.