FEZrs is an open-source Python package for geospatial multispectral image processing and remote sensing. It computes spectral indices, filters, image enhancement, texture, PCA, clustering, change detection and SVM classification from GeoTIFF bands, and writes figures and georeferenced GeoTIFFs.
FEZrs is an open-source Python package for geospatial multispectral image processing and remote sensing. It brings indices, filters, enhancement, texture, PCA, clustering, change detection and SVM classification into a single, consistent framework for research and teaching.
Every calculator inherits one base class, so an NDVI, a Sobel filter and a K-Means classification are called the same way. No plotting or raster I/O code is needed.
Import a calculator
Pick the operation you need from the fezrs package.
Point it at your bands
Pass the GeoTIFF band paths the method requires, such as NIR and red.
Export the result
execute() writes a publication-ready figure; to_raster() then writes a georeferenced GeoTIFF with the source CRS and transform.
from fezrs import NDVICalculator
ndvi = NDVICalculator(nir_path="B5.tif", red_path="B4.tif")
ndvi.execute(output_path="./outputs")
# Optional: full-precision GeoTIFF for GIS
ndvi.to_raster(output_path="./outputs/ndvi.tif")$ pip install fezrs Thirty-seven calculators in five groups, as listed in the FEZrs paper. Each name maps to a class, so NDVI is NDVICalculator.
Satellite-specific helpers (Landsat 8 and GeoEye) are also included.
Built on
Every image below was produced by FEZrs from multispectral scenes in the project's example folder. Select an operation to view its output.
GaussianCalculatorGaussian smoothing suppresses high-frequency noise while keeping broad spatial structure.Drag the handle to move between the two results.
SobelGaussian
K-MeansHSVEach module has a page in FEZdocs with its parameters, band requirements and examples.
NDVI, NDWI, SAVI, AFRI, BI and UI computed from reflectance bands.
Gaussian, mean and median smoothing, and Laplacian and Sobel edge detection.
Histogram and adaptive equalization, and gamma, log and sigmoid adjustment.
Compare images of the same area at two dates to map burns and other change.
K-Means: unsupervised clustering of pixels into k spectral classes.
Support vector machine classification trained on labelled samples.
Principal component analysis: uncorrelated components ordered by variance.
Texture statistics from the gray-level co-occurrence matrix.
Hue–saturation–value transforms that separate color from brightness.
Reflectance plotted against wavelength for a pixel or region.
Merge adjacent rasters into one continuous georeferenced image.
Read single-band products such as Landsat 8 and multi-band products such as GeoEye.
FEZrs is an open-source Python package for geospatial multispectral image processing and remote sensing. Its 37 operations share one calling pattern: create a calculator with the band paths, call execute() for a figure, and to_raster() for a GeoTIFF.
Run pip install fezrs, or install it from the FEZtool channel on Anaconda with conda install -c FEZtool fezrs or mamba install FEZtool::fezrs. FEZrs needs Python 3.11 or later.
NDVI, NDWI, SAVI, AFRI, BI (bare-soil index) and UI (urban index). Each is a calculator class, so NDVI is NDVICalculator.
Yes. execute() saves a figure of the result; to_raster() then writes the full-precision array as a GeoTIFF with the source CRS and affine transform, ready for GIS software.
Yes. FEZrs is released under the Apache-2.0 license, and its documentation under CC BY 4.0.
Cite the archived release on Zenodo, doi:10.5281/zenodo.14938038. The software paper is under review at the Journal of Open Source Software; read the manuscript.