Problem

I've been using ImageJ's numerical propagation plugin written in Java, The visual results from running Angular spectrum using the same parameters/measurements differ substantially between pyDHM and ImageJ Numerical Prop. The ImageJ output appears to be qualitatively more accurate.

Input Image:

ImageJ Numerical Propagation (Angular Spectrum) Results

Input Parameters:

wavelength:  409 [nm]  
distance:  1.1  [mm]  
input width:  3.85 [mm]
input height:  2.76 [mm]  

Visual Output

pyDHM Numerical Propagation(Angular Spectrum) Results

Input Parameters

wavelength:  409 [nm]  
distance:  1.1  [mm]  
input width:  3.85 [mm]
input height:  2.76 [mm]  

Note: All mm parameters are divided by 1000 to bring to meters.
Note: wavelength is divided by 10^9 (otherwise no visual results) So 409 --> 409/(10.0^9)

Visual Output

Code:

mm_to_m = lambda x: x/(10.0**3)
nm_to_m = lambda x: x/(10.0**9)
params = dict(wavelength = nm_to_m(409), 
                       z = mm_to_m(1.01), 
                      dx = mm_to_m(3.85), 
                      dy = mm_to_m(2.76))

image = utilities.imageRead("test_data/plankton.jpg")   

# filter is not applied (Nor is it applied in other Numerical prop tools)  
output = angularSpectrum(image, **params)    

enhanced_output = utilities.intensity(output, False)     

General Question about filters

If I did not wish to apply a visual filter or restrict the ROI and wanted numerical propagation on the full image, Are there any transforms within the sfr, or sfmr that I'm missing outside of it?

The spatial filtering function for the phase compensation methods is taking too long. This function needs to be optmized.

The criterium must be checked and improved in the thresholding-and-summation metric (the function that allows quantifying the degree of compensation of a phase map).

An issue has been detected for the spatial filtering function when non-square images of size MxN are input. Would you please check this issue and solve it? To validate, employ non-squared DHM holograms.

Angular spectrum is described as:

angularSpectrum(field, z, wavelength, dx, dy)

Function to propagate a complex distribution using the angular spectrum approach. field is the input complex wavefield to be propagated. The distance to propagate the input wavefield is represented by z. wavelength is the wavelength of the illumination source used to record the hologram. Finally, dx and dy are the pixel size for the input and output planes along the x- and y- directions.

Is this z, dx and dy distance in meters? millimeters?

Both in the phase compensation and in the phase-shifting methods the search of the brightest point of the +1 diffraction order inside the DHM spectrums is working rarely. Please check and improve.