I'm researching some aspects of Fourier Ptychography, where one illuminates a sample with an LED. Each LED causes a shifted spectrum in fourier space which can be stitched to a high resolution image. Actually I don't need high resolution, just an estimate of objects phase.
So my question: Is it possible to get an estimate of objects phase by capturing - lets say - 9 illumination angles (centre + 8 at NA=0.4.. symmetrically around the centre) and retrieve the Phase by the same procedure as the paper in the link describes?
My results so far are not great. Any ideas?
http://www.nature.com/nphoton/journal/v7/n9/full/nphoton.2013.187.html
Three things I can think of:
1. Do you have enough overlap for each illumination angle? It's critical for the algorithm to work.
2. 9 illumination angles might not be enough for ptychography to work. You can try more and see if it makes any difference.
3. Check the algorithm. Do some simulations first. Make sure the codes work.
Three things I can think of:
1. Do you have enough overlap for each illumination angle? It's critical for the algorithm to work.
2. 9 illumination angles might not be enough for ptychography to work. You can try more and see if it makes any difference.
3. Check the algorithm. Do some simulations first. Make sure the codes work.
PS: In case it helps, here is one of my papers about Fourier Ptychography.
Thank you very much for your advice! Actually the overlap of the images spectrum was one thing I changed in my settup just before you answered :). Thank you for the paper as well!
I'm using a LCD which changes the illumination-angle, so it's easy to change the max. NA used in the off-axis illumination. I've read somewhere that the min. overlap should be around 60-70% which is fullfilled now. Adding more illumination angles would probably result in a better phase estimate.
I'm imaging microbeads d=6µm in water at 9 angles (center, 8*30° spacing at NA=.4) at 530nm with a 63x 0.85 objective lens. The result can be seen in the attachement 1.jpg (left: one raw image, middle unwrapped phase, right: reconstructed amplitude).
Do you have an idea what could cause the artefacts around the sphere? In my simulated data (same setup but numerical simulation) I see the same effect. (attachement 2.jpg) Is it due to the effect of not enough overlap?
Thank you very much!
Ptychography imaging beads using visible light could be problematic. The reason is that the beads could act like strong lenses, which might be the case here judging from the raw image (these bright spots), although you mentioned the beads are in water to match the refractive index. If the beads really act like lenses, it will break down the multiplicative approximation (the exit wave is given by the multiplication of the probe and object transmission function) and ptychography won't work. So I suggest to do a test experiment with a microscope slide and make sure it work. Despite all these, the simulation should work perfectly. So you might want to double check your codes.
sir, Please do respond.
In fourier Ptychography how the filietered high resolution spectrum( the specturm of intialized high resolution) is converted to low resolution image?
Do we need to sub sample the spectrum? or
Do we need to collect the pixels around the filtered region? I am not clear with this step
In my understanding and from what I've read in the publication mentioned above you only need the initial spectra as something like a wild-cart. You could also start with a zero-valued spectra. You're iteratevly replacing its values by the low resolution images measured at different angles. This angles give you the actual position in frequency-space.
You can see the procedure here:
http://www.nature.com/nphoton/journal/v7/n9/fig_tab/nphoton.2013.187_F1.html
I have a small doubt about my Simulation of FPM
I filtered the high resolution image spectrum.I inverse transformed the low resolution image then I got an output(Please see the attached) Is this correct.
Should I have to take care of dimensions?
There is no attachment. Did you look up some example codes on Github?
Sorry for the late reply. I can see an image now, but I can hardly identify anything.. You definitly should take care of the dimensions! You're acquiring a raw-image with effective pixel-size and aperture size. When running your FPM - Reconstruction you need these parameters for the correct placement of the sub-spectrum. Otherwise it wouldn't work at all! It's highly sensitv to the angle of illumnation which is responsable for the shift in fourier-space...What kind of object did you acquire above?
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Waves