Create test_remove_background_noise.py

tests/test_remove_background_noise.py

@@ -0,0 +1,100 @@
+
+import sys
+sys.path.append('.')
+sys.path.append('./src')
+
+import warnings
+warnings.filterwarnings("ignore")
+
+import logging
+logger = logging.getLogger('')
+
+from pathlib import Path
+import numpy as np
+import cupy as cp
+from tifffile import imwrite
+from src.synthetic import SyntheticPSF
+from src.preprocessing import na_and_background_filter, dog
+
+
+import imageio
+import pytest
+
+@pytest.mark.run(order=1)
+def test_filter(kargs):
+    """
+    Will generate the filter response for removing background filters (dog and na_and_background_filter)
+
+    """
+    high_sigma = 6  # sets the low frequency cutoff.
+
+    # Create lattice SyntheticPSF so we can get the NA Mask
+    samplepsfgen = SyntheticPSF(
+        order='ansi',
+        n_modes=kargs['num_modes'],
+        distribution='mixed',
+        mode_weights='pyramid',
+        signed=True,
+        rotate=True,
+        psf_type='../lattice/YuMB_NAlattice0p35_NAAnnulusMax0p40_NAsigma0p1.mat',
+        lam_detection=kargs['wavelength'],
+        psf_shape=[64, 64, 64],
+        x_voxel_size=kargs['lateral_voxel_size'],
+        y_voxel_size=kargs['lateral_voxel_size'],
+        z_voxel_size=kargs['axial_voxel_size'],
+    )
+
+    # real space image = a single voxel at the center of a volume.  This will have a uniform otf.
+    realsp = cp.zeros(samplepsfgen.psf_shape)   # [64, 64, 64]
+    center = np.array(realsp.shape)//2
+    realsp[center[2], center[1], center[0]] = 1
+    realsp -= np.mean(realsp)                   # set DC frequency to zero so it doesn't overload otf.
+    fourier = samplepsfgen.fft(realsp)
+
+    base_folder = Path("./filter")
+    base_folder.mkdir(exist_ok=True)
+    print(f"\nUsing {high_sigma=}\n")
+    print(f"Outputing results to : {base_folder.resolve()}\n")
+    imwrite(f'{base_folder}/fourier.tif', np.abs(cp.asnumpy(fourier)).astype(np.float32))
+    imwrite(f'{base_folder}/realsp.tif', np.abs(cp.asnumpy(realsp)).astype(np.float32))
+
+    """
+    # checking this step by step 
+    fourier = samplepsfgen.fft(realsp)
+    fourier[samplepsfgen.na_mask() == 0] = 0
+    im1 = samplepsfgen.ifft(fourier)
+    FFTfiltered_realsp = im1
+    
+    fourier = np.fft.fftshift(np.fft.fftn(np.fft.ifftshift(realsp)))
+    fourier[samplepsfgen.na_mask() == 0] = 0
+    FFTfiltered_realsp = samplepsfgen.ifft(fourier) #
+    FFTfiltered_realsp = np.abs(np.fft.fftshift(np.fft.ifftn(np.fft.ifftshift(fourier))))
+    """
+
+    # filter the real space image. Remove DC.
+    FFTfiltered_realsp = na_and_background_filter(realsp, low_sigma=0.7, high_sigma=high_sigma, samplepsfgen=samplepsfgen)
+    FFTfiltered_realsp -= np.mean(FFTfiltered_realsp)
+
+    dogfiltered_realsp = dog(realsp, low_sigma=0.7, high_sigma=high_sigma)
+    dogfiltered_realsp -= np.mean(dogfiltered_realsp)
+
+    imwrite(f'{base_folder}/FFTfiltered_realsp.tif', np.abs(cp.asnumpy(FFTfiltered_realsp)).astype(np.float32))
+    imwrite(f'{base_folder}/dogfiltered_realsp.tif', np.abs(cp.asnumpy(dogfiltered_realsp)).astype(np.float32))
+
+    FFTfourier = samplepsfgen.fft(FFTfiltered_realsp)
+    dogfourier = samplepsfgen.fft(dogfiltered_realsp)
+
+    FFTfiltered_otf = np.abs(cp.asnumpy(FFTfourier)).astype(np.float32)
+    dogfiltered_otf = np.abs(cp.asnumpy(dogfourier)).astype(np.float32)
+
+    imwrite(f'{base_folder}/FFTfiltered_otf.tif', FFTfiltered_otf)
+    imwrite(f'{base_folder}/dogfiltered_otf.tif', dogfiltered_otf)
+
+    # Save principle planes
+    imageio.imsave(f'{base_folder}/FFTfiltered_otf_XY.png', FFTfiltered_otf[center[2],:,:])
+    imageio.imsave(f'{base_folder}/dogfiltered_otf_XY.png', dogfiltered_otf[center[2],:,:])
+    imageio.imsave(f'{base_folder}/FFTfiltered_otf_XZ.png', FFTfiltered_otf[:,center[1],:])
+    imageio.imsave(f'{base_folder}/dogfiltered_otf_XZ.png', dogfiltered_otf[:,center[1],:])
+    imageio.imsave(f'{base_folder}/FFTfiltered_otf_YZ.png', np.transpose(FFTfiltered_otf[:,:,center[0]]))
+    imageio.imsave(f'{base_folder}/dogfiltered_otf_YZ.png', np.transpose(dogfiltered_otf[:,:,center[0]]))
+