Awesome

SubpixelAlignment

Implementation of image alignment through phase correlation in Fourier space for pixel- and subpixel-bias.

Dataset

The data is located in the data directory. It contains two datasets: saturn and sentinel2. Each dataset consists of two .tiff images of the one region, separated in time.

Code

The code is located in the code directory. It contains the scripts for dataset uploading (dataset.py), drawing (plot.py), and detection of the offset between images via phase correlation.

Note: if the offset is assumed to be on the order of pixels, the parameter upsample_factor should be equal to 1 (saturn dataset case); otherwise, if the shift between images is observed on the subpixel level, the parameter upsample_factor should be more than 1 (typically 3-10; sentinel2 dataset case).

The example of using the code is presented in the Jupyter Notebook (data/example.ipynb).

Requirements

• numpy==1.18.5
• rasterio==1.1.5
• scikit-image==0.17.2
• scipy==1.4.1

(see requirements.txt for the more details)