Meeting Summary

[chengkunli96]

The code of the Angular Spectrum Function has been implemented. And for testing its correctness, I designed the Fraunhofer Diffraction (I make a Gaussian beam incident on a plane that has a circle and rectangle aperture in its centre respectively) to see the result whether it is the same as our prediction.

Next, I should do an extended experiment to check its correctness. Using this function to propagate a complex image at z distance. we will get a new complex image. And then, when you try to propagate this new image back at -z distance, you should get the same image as the original one. Secondly, do some experiment to find the propagation distance range for which this function works. And finally, book a project meeting when all above have done.

--18/03/20121

[chengkunli96]

Add padding which is 2 times (2x for the input image) in my code. Because that the phase-object some times is very complex, we want to track all the information at the incident plane.

And the next thing is to implement the angular spectrum function with tilted planes. The references can be found here (https://www.osapublishing.org/josaa/abstract.cfm?uri=josaa-20-9-1755) and (https://www.osapublishing.org/oe/fulltext.cfm?uri=oe-13-24-9935&id=86288).

--01/04/2021

[chengkunli96]

The rotated angular spectrum function (for computing optical diffraction on the tilted plane) has been implemented. However, there seems some question on the interpolation step. We can change the input image to check the correction, like setting the input phase as random values in between zero and two pi. (We can also skip this step firstly)

For the next step of the project. Firstly, create a dataset that has multiple input images that are diffracted (both phase and amplitude) and output images that are tilted (both phase and amplitude) (Tilt: 10 degrees). Then, use the generated dataset to train a U-Net to implement the forward beam propagation on the tilted plane.