Skip to main content

Category comparison

Scientific image analysis software compared: choosing between Fiji, QuPath, CellProfiler, ilastik, napari and OpenCV

Six open-source image analysis tools for research compared by image type and by whether you write code: what each is good for, what it is not suited for, and what a measurement still needs before it goes into a paper.

Page updated:
2026-11-16
Item-by-item check against official documentation:
not done yet
Uniform test:
A comparison of documentation; no uniform test has been run yet

Before choosing image analysis software, answer three questions: what kind of images you have (ordinary microscope images, whole pathology slides, 3D or time series), whether you have a few images or thousands, and whether you are willing to write code.

All six tools are open source and are more often combined than swapped for one another. No comparative test was run for this page; the "good for / not suited for" lines are an editorial summary that has not yet been checked item by item against each tool's official documentation.

Start from what you need to do

  • A few images; measure length, area or intensity by hand or semi-automatically

    Tools to considerFiji / ImageJ

    A graphical program with many plugins and plenty of documentation. Steps can be recorded as a macro and repeated later.

  • Cell detection and region annotation on whole slide images

    Tools to considerQuPath

    QuPath is built for very large slide images: you can browse them smoothly and run detection and classification on them.

  • The same quantitative analysis on hundreds or thousands of images, without code

    Tools to considerCellProfiler

    You assemble the steps into a pipeline and run it on the whole set, with the same parameters for every image.

  • A threshold cannot separate objects from background; you want to teach the software by marking examples

    Tools to considerilastik

    You paint a few labels, and it trains a classifier and shows the result immediately. Its output is often passed to Fiji or CellProfiler for measurement.

  • Analyse 3D, multichannel or time-lapse images in Python, with viewing and annotation

    Tools to considernapari

    napari is a multidimensional image viewer for Python that works directly with NumPy, scikit-image and similar libraries.

  • Build image measurement into your own program or automated pipeline

    Tools to considerOpenCV

    OpenCV is a library with filtering, segmentation, contours and geometric measurement, suited to repeatable batch processing.

Basic facts side by side

ToolCostLicencePlatformsRecord checked
Fiji / ImageJCost not verifiedGPL-3.0Windows · macOS · Linuxnot checked
QuPathCost not verifiedGPL-3.0Windows · macOS · Linuxnot checked
CellProfilerCost not verifiedBSD-3-ClauseWindows · macOS · Linuxnot checked
ilastikCost not verifiedGPL-2.0Windows · macOS · Linuxnot checked
napariCost not verifiedBSD-3-ClauseWindows · macOS · Linuxnot checked
OpenCVCost not verifiedApache-2.0Windows · macOS · Linux · Android · iOSnot checked

What each tool is good for and not suited for

Open-source scientific image analysis platform for SEM, TEM, AFM and microscopy image measurement.

Good for
General processing and measurement of microscope images, with many plugins and macro-based batch processing.
Not suited for / note
Very large images such as whole slides; complex procedures done only by clicking are hard to repeat.

Open-source digital pathology and bioimage analysis software for large tissue section images.

Good for
Viewing, annotating, detecting and classifying cells on digital pathology slides.
Not suited for / note
It is not a general image-processing tool; routine measurement of ordinary microscope images is more direct in Fiji.

Free, open-source image analysis software from the Broad Institute for high-throughput cell image analysis.

Good for
Building an analysis pipeline without code and quantifying large image sets consistently.
Not suited for / note
Exploring single images is less flexible than in Fiji; the pipeline has to be tuned on a few images first.

Interactive machine-learning image segmentation tool that lets non-programmers perform pixel classification.

Good for
Training pixel classification by interactive labelling, for images a threshold cannot segment.
Not suited for / note
Its own measurement functions are limited; with few or unrepresentative labels the result is unreliable.

Open-source tool in the Python ecosystem for viewing, annotating and analysing multidimensional scientific images.

Good for
Viewing and annotating multidimensional images in Python alongside your own analysis code.
Not suited for / note
It needs a Python environment and programming. It is a viewer and platform; the analysis comes from plugins or your own code.

Open-source computer vision library for image processing, video analysis, object detection and vision algorithm development.

Good for
Repeatable image processing and geometric measurement inside a program.
Not suited for / note
No graphical interface; microscope file formats are supported less directly than in Fiji or QuPath.

What this page does not answer

  • Software that outputs a number has not thereby validated the measurement. Pixel calibration, threshold and segmentation method all change the result; see the article on validating image measurement on this site.
  • This page does not compare speed or accuracy; both depend on the images and the parameters.
  • The ZIRULAB tool library has two records for Fiji ("Fiji / ImageJ" and "ImageJ / Fiji"); they are the same software, and only one is listed here.

Related articles and workflows

Articles

Workflows

Where the information comes from

Cost, licence and platforms come from the ZIRULAB tool records; the table gives the date each record was checked, and "not checked" where there is no check record. The "good for / not suited for" lines are an editorial summary that has not yet been checked item by item against each tool's official documentation.

ZIRULAB ran no comparative test for this page; it contains no test scores and no ranking.

How content is verified