search

fiji / SPIM_Registration

Deprecated, instead check out: https://github.com/preibischLab/multiview-reconstruction & https://github.com/PreibischLab/BigStitcher
GNU General Public License v2.0
23 stars 22 forks source link

Very Strong artifacts #117

Open lacan opened 7 years ago

lacan commented 7 years ago

Dataset: 5 Views of about 400 slices each.

2 Channel images, with one channel containing the beads ( 561nm ), the other the sample ( 488nm ).

Registration done on 200nm beads, then duplicated to other channel.

Voxel Size: 320nm in xy and 1000nm in z

Deconvolution done using the beads of channel 561 for both channels on CUDA 10 iterations, all defaults except for blocksize of 1024x1204x256, effective block size around 972x970x180

At the end we get terrible artifacts on the images and have no idea where these could be from.

image

I would be very grateful for some insight on this.

Deconvolution on smaller images taken with the same modality seem to be unaffected.

lacan commented 7 years ago

I can submit the HDF5 Dataset if needed (~8GB)

stevenedwards commented 7 years ago

I've also seen similar artefacts when using 1024x1024x768 block size on CUDA. Using smaller block sizes (512x512x512) solved the issue (as a quick fix).