Table 1
Deconvolution results for the SL1M algorithm for a variety of regularisation parameters (λ) run against 37 800 visibilities simulated from a test image of 50 point sources on a 1024 × 1024 pixel grid.
λ | Iter | L2 term | L1 term | Error | Error |
(rms) | (max) | ||||
|
|||||
10-2 | 7000 | 8.3 × 10-3 | 1448 | 8.1 × 10-3 | 1.61 |
10-1 | 7000 | 4.7 × 10-2 | 655.5 | 4.6 × 10-3 | 0.90 |
1 | 4992 | 8.6 × 10-1 | 61.58 | 5.5 × 10-6 | 4.3 × 10-4 |
101 | 1860 | 4.8 | 60.62 | 3.9 × 10-6 | 1.1 × 10-3 |
102 | 1367 | 9.8 | 60.60 | 1.2 × 10-5 | 4.2 × 10-3 |
103 | 454 | 460 | 59.90 | 1.1 × 10-4 | 2.8 × 10-2 |
Notes. No noise has been added to the visibilities. The runs were limited to 7000 iterations or a change in total error of 1 part in 106. The error terms for Eq. (10) are shown separately in the third and fourth columns. The reconstruction errors are based on comparison with the original image. Note that the true L1 norm of the input data was 60.5996.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.