About Dragonfly's Segmentation Tools
Segmentation, also known as classification or labelization, is the process of partitioning an image into multiple segments or sets of voxels that share certain characteristics. The main goal of segmentation is to simplify or change the representation of an image into something that is more meaningful and easier to analyze. Another objective of segmentation is to organize an image into higher-level units, such as meshes, graphs, and vector fields, that are more efficient for visualization and further analysis.
In simple cases, environments may be well enough controlled so that the segmentation process reliably extracts only the parts that need to be analyzed further. In complex cases in which boundaries are indistinct, such as missing edges or a lack of contrast between foreground and background regions, segmentation can be more difficult. In either case it is important to understand that:
- There is no universally applicable technique that will work for all types of images.
- No segmentation technique is perfect.
- Post-processing, by smoothing noisy images or enhancing edges, can accelerate some segmentation tasks (see Image Filtering).
- If you are segmenting large volumes, it may be helpful to simplify the volume before segmentation (see Cropping Datasets).
The following tools and options are available in Dragonfly for segmenting image data.
Thresholding
The options in the Range box on the ROI Tools panel allow you to define an intensity domain of image data values and then apply the selected range as a threshold segmentation (see Thresholding). You can also add or remove a range from a selected ROI. Intensity domains are also applicable to other tools, such as the morphological operators and ROI Painter tools.
ROI Painter Tools
The tools on the ROI Painter panel are used for manual segmentation and for editing regions of interest in 2D and 3D views (see ROI Painter).
Histographic Segmentation
With this tool, you can quickly segment a dataset by selecting clustered data values, which will be propagated to all data points that meet the selection criteria. Regions of interest created in this manner can either be exported or expanded through a Watershed algorithm to fully segment the selected dataset (see Histographic Segmentation).
Active Contours
The Active Contour workflow begins with adding a series of paths to the 2D views of volumetric image data, fitting the closed splines (known as snakes) to object boundaries, and then generating a mesh that fully describes the surface of the targeted feature of interest (see Active Contour).
Segmentation Wizard
Dragonfly’s Segmentation Wizard provides an easy-to-use, guided workflow for implementing powerful deep learning and classical machine learning segmentation of multi-dimensional images (see Segmentation Wizard).
Machine Learning Segmentation
This advanced machine learning plug-in for image segmentation, provides an opportunity to train a classifier within a limited sample in an image so that it will learn how to segment the pixels of the whole dataset or other similar datasets (see Machine Learning Segmentation).
Deep Learning
With Dragonfly's Deep Learning Tool even non-experts in image processing and artificial intelligence can create robust and reproducible segmentation results by training a deep model for semantic segmentation (see Deep Learning).
