tgv_pycuda

tgv_pycuda implements primal-dual algorithms for imaging problems regularized by total variation (TV) and second-order total generalized variation (TGV). The Python implementation uses PyCUDA for GPU acceleration and requires a CUDA-capable GPU and a working CUDA installation.

The repository includes algorithms, tests, and examples for denoising, deblurring, zooming, dequantization, and compressive imaging. A guided Jupyter notebook reproduces figures and numerical experiments from the associated publication on recovering piecewise smooth multichannel images.