It's parallelized implementation of multi-scale image decomposition algorithms. Current impl. includes CUDA based Lift-DWT and Mallat-DWT. Additionally there is an initial impl. of Mallat-DWT parallelized with OpenMP for multi-core CPUs. Finally, current cuda based software accelerates data transfers on Jetson TX2.
- Asynchronous cuda-streams.
- Pipelined-overlap I/O transfers with a computation kernels.
- C++11 compiler
- CMake 3.x
- CUDA libs and headers
- git clone /dst
- cd /dst
- mkdir build
- cd build
- cmake ..
- make
-
Real-time fusion of visible and thermal infrared images in surveillance applications on SoC hardware. https://doi.org/10.1117/12.2325391
-
Implementation of the DWT in a GPU through a Register-based Strategy. https://doi.org/10.1109/TPDS.2014.2384047