Peaks not detected in registration step -

detecting peaks.txt (11.2 KB)

Hello everyone,

Hoping to get some help here, please.

I am running a reconstruction of 7 image stacks (of beads in agarose imaged on an OpenSpim setup) using the multiview reconstruction plugin in Fiji.

At the point of detecting interest points, after defining a suitable threshold for the difference of a Gaussian, I noticed peaks are only detected for the 7th stack, while a message of ‘no peaks detected’ is thrown up for other stacks. I was wondering if someone can help on how to move past this. I have attached a file with the error message.

The stacks are tiff files and are more than 4GB (except the 7th stack). Could this be the problem? I loaded them using the Image stacks (ImageJ opener)… when defining the XML file.

I had problems loading these images when using the SPIm bead based registration plugin. So I am guessing the challenge might be the known problem of imageJ with tiffs that are more than 4GB.


@StephanPreibisch @schmiedc @hoerldavid @haesleinhuepf

did you check that these views in the hdf5 did not contain anything?
Maybe try to use the Image Stacks (LOCI Bioformats) for opening the files and resaving as hdf5.

Hi @schmiedc

I have tried several options over the last few days. The solution is just to limit each stack to 4GB.


That is unfortunately sub-optimal if the limitation is really just the file dimensions.

However, in my experience it is often better and sufficient to reduce the sampling per view (either in xy by using a tighter ROI, or in z by increasing the step size). Makes all things down stream easier with little compromise on the data quality. Since the fusion/deconvolution of the multiple views results in the isotropic resolution.

As far as I know there where some efforts to directly save into hdf5 in the OpenSPIM. @hkmoon spearheaded this direction. Maybe @Johannes_Girstmair followed up on this.

Hi Chris,

Exactly what I meant. reducing sampling per view to reduce size. I tried saving as HDF5 and still didn’t work.

Thank you.