These are the requirements of MIB 3. For MIB 2 and MIB 1 see the previous versions.

MIB 3 runs as a MATLAB program under Windows, Linux and macOS installations of MATLAB, or as a compiled (standalone) version for various operating systems that does not require a MATLAB license for academic research, see Downloads.
The standalone MIB requires the MATLAB Runtime, which is installed automatically during installation.
A 64-bit operating system with a sufficient amount of memory is recommended. For DeepMIB a CUDA-capable NVIDIA GPU is required.

only for the MATLAB version of MIB

• MATLAB R2025a or newer (R2026a is recommended). For older MATLAB releases use MIB 2
• MATLAB R2026b and newer no longer include Java. MIB works without it, but Bio-Formats, Fiji, Imaris and OMERO need Java: how to enable Java

only for the MATLAB version of MIB; the standalone version includes everything it needs

• Image Processing Toolbox is required
• Deep Learning Toolbox (required for DeepMIB) for training of 2D and 3D networks and segmentation of datasets
• Computer Vision Toolbox (required for DeepMIB), also used for feature-based alignment and for adding text to snapshots
• Parallel Computing Toolbox (required for DeepMIB and other tools that benefit from parallel processing) for GPU computing and for speeding up computations with parallel workers
• Statistics and Machine Learning Toolbox (optional) for alternatives to the Random Forest classifier
• Optimization Toolbox (optional) for alternatives to the Random Forest classifier

Install from MATLAB->Home tab->Add-Ons->Get Add-Ons:

• Computer Vision Toolbox Model for SOLOv2 Instance Segmentation for instance segmentation with the SOLOv2 network in DeepMIB
• Deep Learning Toolbox Converter for ONNX Model Format for exporting trained DeepMIB networks to the ONNX format
• Deep Learning Toolbox Converter for TensorFlow Models for exporting trained DeepMIB networks to the TensorFlow format

The Bio-Formats library brings support for multiple proprietary microscopy image formats. It is included with MIB, no installation is needed. To use it, select the BioFormats reader in the Directory Contents panel.

• With MATLAB R2026b and newer, Bio-Formats needs Java: how to enable Java
• Opening large files is faster from the second time on, when a temporary directory for the Bio-Formats Memoizer is set in Home tab->Preferences->External directories

BM3D and BM4D filters provide exceptional filtering quality. The filters are not supplied with MIB due to license limitations (non-profit, non-commercial use only), but can be used together with MIB when installed on the system.

only for the MATLAB version of MIB:

• Check the license limitations
• Download BM3D MATLAB 4.0.3 (bm3d_matlab_package_4.0.3.zip) or newer from https://webpages.tuni.fi/foi/GCF-BM3D/; it already contains BM4D, a separate BM4D download is not needed
• Unzip to your scripts folder, for example c:\MATLAB\Scripts\bm3d_4.0.3\
• Open MIB preferences: Home tab->Preferences->External directories
• Select the unzipped folder in BM3D installation directory
• Restart MIB
• After restart the filter is available from Image tab->Image filters->Edge-preserving filtering->BMxD

Note! MIB 3 does not work with the legacy BM3D 2.01.

The Fiji Connect panel allows exchange of datasets and models between MIB and Fiji, using MIJ (included with MIB). Both the MATLAB and the standalone versions of MIB can use Fiji.

• Download Fiji and unzip it, for example to c:\Tools\Fiji.app\
• Start Fiji once and update it: Fiji->Menu->Help->Update...
• Open MIB preferences: Home tab->Preferences->External directories
• Select the Fiji folder in Fiji installation directory
• Restart MIB. MIB adds the Fiji libraries itself, there is no need to edit the MATLAB path
• With MATLAB R2026b and newer, Fiji needs Java: how to enable Java

See the Fiji Connect panel in the user guide for details.

Microscopy Image Browser can be used together with Imaris. This functionality is achieved using IceImarisConnector written by Aaron C. Ponti, ETH Zurich.
Note! The connection to Imaris was converted from MIB 2 but has not been tested with MIB 3 yet.

• Install Imaris and ImarisXT
• Open MIB preferences: Home tab->Preferences->External directories
• Select the Imaris installation folder in Imaris installation directory
• Restart MIB
• With MATLAB R2026b and newer, Imaris needs Java: how to enable Java

The Membrane Click Tracker tool uses compiled Accurate Fast Marching functions. They are already compiled for Windows, Linux and macOS (Intel and Apple silicon).
For other systems, compile them from the MATLAB command window:

• Change directory to mib\external\FastMarching\, where mib is the MIB folder, for example c:\MATLAB\MIB3\mib
• Run compile_c_files

Note! The connection to an OMERO server is not implemented in MIB 3 yet. To work with OMERO, use MIB 2, see its OMERO requirements.

The Random Forest Classifier uses randomforest-matlab by Abhishek Jaiantilal, which is already compiled for Windows.
For other systems the files have to be compiled manually, see mib\external\RandomForest\RF_Class_C\README.txt and mib\external\RandomForest\RF_Reg_C\README.txt.

MIB uses its own function for saving data in the NRRD format, but relies on Projects:MATLABSlicerExampleModule by John Melonakos for reading it.
The reader is already compiled for Windows; for other systems see mib\external\nrrd\compilethis.m.

Segment-anything (SAM) requires a Python environment, see the dedicated page with the installation details: Segment-anything model.

The brush tool can be used to select not individual pixels but rather groups of pixels (superpixels). This functionality is implemented using the SLIC (Simple Linear Iterative Clustering) algorithm written by Radhakrishna Achanta et al., 2015. In addition, the SLIC superpixels and supervoxels are used for the Graph-cut segmentation and Classifier.
For the Graph-cut segmentation MIB is utilizing maxflow 2.22 written by Yuri Boykov and Vladimir Kolmogorov.

These functions are already compiled for Windows, Linux and macOS (Intel and Apple silicon). For other systems, compile them from the MATLAB command window:

• Change directory to mib\external\Supervoxels\, where mib is the MIB folder, for example c:\MATLAB\MIB3\mib
• Run compile_c_files

only for the MATLAB version of MIB
Reading whole-slide images with the OpenSlide reader of the Directory Contents panel requires the MATLAB support package Medical Imaging Toolbox Interface for Whole Slide Imaging File Reader: MATLAB->Home tab->Add-Ons->Get Add-Ons, then search for Whole Slide Imaging File Reader.

MIB 3 reads and writes Zarr v2 and v3 datasets (local and remote, images and models) with the included zarr-matlab engine; no installation is needed.
Python is optional: the zarr-python library is only used when it is selected in Home tab->Preferences->Zarr library. It requires a Python environment with the zarr and numpy packages (plus aiohttp and requests for remote datasets), set in Home tab->Preferences->External directories->Python installation path; see Zarr format for the installation of the Python environment.