rafael-carrillo
commented
10 years ago Hello,
The problem in the script Experiment3.m is the sampling profile in the Fourier domain and not the algorithm itself. Since apriori we assume we are reconstructing real signals the sampling profile has been only defined in half of the Fourier plane and therefore relying on conjugate symmetry (or a real or positive constraint). You can check this by plotting the logic matrix mask.
To work with complex signals you would need to generate a sampling profile that covers the entire Fourier plane. You can do this by just sampling uniformly at random or by modifying the existing routine sopt_mltb_vdsmask (lines 57 to 62) to avoid setting half of the Fourier plane to zero.
Thanks a lot.
On Jul 17, 2014, at 6:13 PM, ahmad46 notifications@github.com wrote:
I want to reconstruct non-positive and/or complex images using SARA. For the demo Experiment3.m, if I drop the positivity and reality constraints by setting param.pos_B2 = 0 and param.real_B2 = 0, SARA does not generate good results, even when the undersampling ratio is 1.0.
— Reply to this email directly or view it on GitHub.
I want to reconstruct non-positive and/or complex images using SARA. For the Matlab demo Experiment3.m, if I drop the positivity and reality constraints by setting param.pos_B2 = 0 and param.real_B2 = 0, SARA does not generate good results, even when the undersampling ratio is 1.0.