Classical segmentation methods
Semantic segmentation methods are most suitable if individual objects are either well separated or can easily be separated by post-processing or the area covered by objects is relevant rather than the precise counting and measuring of individual objects.
The methods are as follows:
Global Thresholding
Applies a fixed threshold across the entire image.
Use Case: Segmenting objects with consistent intensity, such as fluorescence images.
Background Subtraction
Employs a rolling ball algorithm to subtract uneven background before applying a global threshold.
Use Case: Segmenting images with varying background intensity.
Variance-based Thresholding
Segments objects based on intensity changes rather than consistent intensity.
Use Case: Segementing objects in brightfield images where the objects are distinguished by intensity variance.
Dynamic Thresholding
Applies local thresholds to handle inhomogeneous backgrounds.
Use case: Segmenting images with uneven illumination.
