Commands Overview

Pipeline commands are the building blocks of an analysis. Each command takes either an image or a set of objects as input and produces either a processed image or an updated set of objects as output.

Command Categories

Category	Colour	Input	Output
Image processing	Grey	Image	Image
Edge & feature detection	Grey	Image	Image
Segmentation	White	Image	Binary mask
Morphology	White	Binary/greyscale image	Image
AI segmentation	White	Image	Binary mask / object instances
Object processing	Green	Objects	Objects / measurements
Measure	Blue	Image + final objects	Measurements
Dataset	-	Objects of all images	Objects / measurements

Note

Not every command sequence is valid. The command picker shows only those commands compatible with the current pipeline state.

Image Processing

Commands for reducing noise and enhancing signal before segmentation.

Command	Purpose
Blur	Box/average blur to reduce noise
Gaussian Blur	Edge-preserving Gaussian blur
Rolling Ball	Background subtraction
Illumination Correction	Remove uneven illumination (vignetting) across the whole image
Rank Filter	Min / median / max neighbourhood filter
Enhance Contrast	Histogram-based contrast enhancement
Color Filter	Retain pixels within an HSV colour range
Intensity Transform	Brightness and contrast adjustment
Median Subtract	Subtract the local median for background removal
Image Math	Combine two images mathematically
Image Cache	Store or retrieve an image from cache slots
Load Image Channel	Load another channel as the current image
Script	Run your own Rhai script, using the other commands

Edge & Feature Detection

Commands for detecting edges, ridges, and local structures.

Command	Purpose
Canny Edge Detection	Gradient-based edge detection with thresholding
Sobel Edge Detection	Fast gradient-magnitude edge detection
Hessian	Second-order structure (blobs and ridges)
Laplacian	Zero-crossing edge detection
Structure Tensor	Local orientation and coherence
Weighted Deviation	Weighted local standard deviation

Segmentation

Commands for separating foreground objects from background.

Command	Purpose
Threshold	Convert greyscale to binary with manual or automatic threshold
Connected Components	Label each connected foreground region
Watershed	Split touching objects using intensity valleys

Morphology

Commands for modifying the shape of binary or greyscale regions.

Command	Purpose
Morphological Transform	Erosion, dilation, opening, closing

AI Segmentation

Commands that run a pretrained deep-learning model for segmentation. Only available in builds with the ai Cargo feature enabled.

Command	Purpose
Stardist	Instance segmentation via star-convex polygons - separates touching objects directly
UNet	Semantic foreground/background mask - pair with Connected Components (+ Watershed) to separate instances
Cellpose	Instance segmentation via flow-field dynamics - handles irregular and overlapping shapes
YOLOv5	Instance segmentation or detection with boxes + masks - supports several object classes in one model
AI Pixel Classifier	Semantic mask from a Random Forest/k-NN/MLP model you train yourself on painted examples

Object Processing

Commands that operate on extracted objects.

Command	Purpose
Fill Holes	Close enclosed background holes in the segmentation map
Fill Object Holes	Fill the holes inside every object, one object at a time
Load Annotated Objects	Bring hand-annotated objects into the pipeline as regular objects
Extract Objects	Convert binary mask regions to segmentation-class objects
Classify Objects	Filter and assign final object classes
AI Object Classifier	Classify objects with a Random Forest/k-NN/MLP model you train yourself on painted examples
Colocalization	Find overlapping objects across classes
Voronoi	Partition space from object centroids
Object Transform	Scale, snap, expand, shrink, or fit an ellipse to objects
Object Math	Boolean set operations (AND, OR, XOR, Subtract) between two object classes
Distance Transform	Measure distances between object pairs
Propagation	Grow a region around each seed object through a channel (e.g. cells around nuclei)
Relate Objects	Link each child object to its parent (e.g. spots to their cell)
Save Image	Write a control image to disk, optionally with selected classes drawn on top

Measure

Measure steps run once the objects are final. They measure the whole image or every object and add the values to the results.

Command	Purpose
Area Occupied	How much of the image the objects of a class cover
Image Intensity	Brightness statistics of a channel over the whole image
Image Quality	Focus, blur, saturation and noise of a channel
Exclude Image	Leave an image out of the results when its measurements say it’s bad
Measure Colocalization	Pearson, Manders and related coefficients of two channels, per image and object
Measure Texture	Haralick texture features, per image and object
Measure Object Shape	Convex area, solidity, Euler number and radii of every object
Measure Object Neighbors	Neighbours, percent touching and closest objects
Measure Distance To Parent	Each child’s distance to its parent (linked by Relate Objects)

Dataset

Dataset steps run once after every image has been analysed, on the stored objects of all images, frames and wells.

Command	Purpose
Simple Tracker	Link objects of a time-lapse into tracks (nearest neighbour)

Modules

Several steps can be bundled into one reusable step with its own parameters - see Modules.

Typical Pipeline Order

Image
  └─ Preprocessing (Blur, Rolling Ball, Illumination Correction, …)
       └─ Segmentation (Threshold → Connected Components → Watershed)
            └─ Fill Holes / Fill Object Holes (optional cleanup)
                 └─ Extract Objects
                      └─ Classify Objects
                           └─ Object Processing (Colocalization, Propagation, Relate Objects, …)
                                └─ Measure (Image Quality, Object Shape, Texture, …)
                                     └─ Save Image