despeckCL

A C++ library for InSAR and PolSAR denoising images using OpenCL with Python bindings. The code is mostly developed and tested on an Intel HD4000, Intel UHD Graphics 620 and a NVIDIA Tesla V100. I also had some success running it on an AMD RX480, but don't hold your breath. To get started have a look at the examples directory, the sphinx documentation or checkout despeckcl.h. If you have any comments, questions or suggestions just write me an email.

Implemented Filters

So far the following filters are implemented:

• Boxcar: The simple boxcar filter
• Goldstein: The Goldstein InSAR filter, Goldstein, R. M.; Werner, C. L., "Radar interferogram filtering for geophysical applications," Geophysical Research Letters,vol. 25, no. 21, pp.4035,4038, 1998
• NL-SAR: A nonlocal filter for SAR, InSAR and PolSAR, Deledalle, C.-A.; Denis, L.; Tupin, F.; Reigber, A.; Jäger, M., "NL-SAR: A Unified Nonlocal Framework for Resolution-Preserving (Pol)(In)SAR Denoising," Geoscience and Remote Sensing, IEEE Transactions on , vol.53, no.4, pp.2021,2038, April 2015
  • currently this implementation uses only a single search window of fixed size in contrast to the original paper, which uses multiple search window sizes.
  • so far only InSAR filtering is supported.
• NL-InSAR: A nonlocal InSAR filter, Deledalle, C.-A.; Denis, L.; Tupin, F., "NL-InSAR: Nonlocal Interferogram Estimation," Geoscience and Remote Sensing, IEEE Transactions on , vol.49, no.4, pp.1441,1452, April 2011

Requirements

1. A compiler with C++14 and at least OpenMP 4.0 support. So far development is done using gcc but I also occasionally test using clang.
2. An OpenCL Runtime:
   • CUDA Toolkit from NVIDIA also has OpenCL support
   • Beignet for Intel GPUs under Linux
   • AMD ROCm for AMD GPUs
3. CMake of at least version 3.12.
4. Boost is used for serialization and deserialization.
5. GNU Scientific Library provides some special functions needed by NL-SAR
6. clFFT AMD's OpenCL FFT library, which also works on NVIDIA and Intel devices. This is used by the Goldstein filter. Is automatically checked out from github and build when running cmake.
7. SWIG for generating Python and Octave bindings (optional)
8. Sphinx for building the documentation (optional)
9. despeckCL uses Google Test for unit testing and Easylogging++ for logging. Google Test is automatically checked out from github when building despeckCL and Easylogging++ is shipped directly with the source.

Building, Testing and Installation

$ git clone https://github.com/gbaier/despeckCL.git

$ cd despeckCL

$ mkdir build

$ cd build

$ cmake ../

The Python version can also be selected manually in case multiple versions are installed.

$ cmake -DPython3_ROOT_DIR=~/anaconda3/ ../

Compilation is straightforward.

$ make -j4

$ make test

$ make install