Direct deconvolution of radio synthesis images using L1 minimisation

Stephen J. Hardy
POBox 6, Pymble BC, 2073 Australia
arXiv:1310.2078 [astro-ph.IM], (8 Oct 2013)

   author={Hardy}, S.~J.},

   title={"{Direct deconvolution of radio synthesis images using L$_{1}$ minimisation}"},





   keywords={Methods: analytical, methods: numerical, techniques: image processing, techniques: interferometric},








   adsnote={Provided by the SAO/NASA Astrophysics Data System}


Download Download (PDF)   View View   Source Source   Source codes Source codes




We introduce an algorithm for the deconvolution of radio synthesis images that accounts for the non-coplanar-baseline effect, allows multiscale reconstruction onto arbitrarily positioned pixel grids, and allows the antenna elements to have directional dependent gains. Using numerical L1-minimisation techniques established in the application of compressive sensing to radio astronomy, we directly solve the deconvolution equation using GPU (graphics processing unit) hardware. This approach relies on an analytic expression for the contribution of a pixel in the image to the observed visibilities, and the well-known expression for Dirac delta function pixels is used along with two new approximations for Gaussian pixels, which allow for multi-scale deconvolution. The algorithm is similar to the CLEAN algorithm in that it fits the reconstructed pixels in the image to the observed visibilities while minimising the total flux; however, unlike CLEAN, it operates on the ungridded visibilities, enforces positivity, and has guaranteed global convergence. The pixels in the image can be arbitrarily distributed and arbitrary gains between each pixel and each antenna element can also be specified. Direct deconvolution of the observed visibilities is shown to be feasible for several deconvolution problems, including a 1 megapixel wide-field image with over 400,000 visibilities. Correctness of the algorithm is shown using synthetic data, and the algorithm shows good image reconstruction performance. Though this algorithm requires significantly more computation than methods based on the CLEAN algorithm, we demonstrate that it is trivially parallelisable across multiple GPUs and potentially can be scaled to GPU clusters. We also demonstrate that a significant speed up is possible through the use of multi-scale analysis using Gaussian pixels.
VN:F [1.9.22_1171]
Rating: 0.0/5 (0 votes cast)

* * *

* * *

Follow us on Twitter

HGPU group

1660 peoples are following HGPU @twitter

Like us on Facebook

HGPU group

334 people like HGPU on Facebook

* * *

Free GPU computing nodes at hgpu.org

Registered users can now run their OpenCL application at hgpu.org. We provide 1 minute of computer time per each run on two nodes with two AMD and one nVidia graphics processing units, correspondingly. There are no restrictions on the number of starts.

The platforms are

Node 1
  • GPU device 0: nVidia GeForce GTX 560 Ti 2GB, 822MHz
  • GPU device 1: AMD/ATI Radeon HD 6970 2GB, 880MHz
  • CPU: AMD Phenom II X6 @ 2.8GHz 1055T
  • RAM: 12GB
  • OS: OpenSUSE 13.1
  • SDK: nVidia CUDA Toolkit 6.5.14, AMD APP SDK 3.0
Node 2
  • GPU device 0: AMD/ATI Radeon HD 7970 3GB, 1000MHz
  • GPU device 1: AMD/ATI Radeon HD 5870 2GB, 850MHz
  • CPU: Intel Core i7-2600 @ 3.4GHz
  • RAM: 16GB
  • OS: OpenSUSE 12.3
  • SDK: AMD APP SDK 3.0

Completed OpenCL project should be uploaded via User dashboard (see instructions and example there), compilation and execution terminal output logs will be provided to the user.

The information send to hgpu.org will be treated according to our Privacy Policy

HGPU group © 2010-2015 hgpu.org

All rights belong to the respective authors

Contact us: