topaz denoise, what are appropriate input formats?

[larsonmattr]

Hi,

I am trying to use topaz denoise with a movie taken in SerialEM (.st). I am using v 0.2.5.

With original .st files, topaz denoise --patch-size 1024 -o Test_denoised/ Test/*.st,
-> There is an error PIL.UnidentifiedImageError: cannot identify image file

When I convert first to .mrc via the EMAN2 e2proc3d.pyprogram, and use then:
topaz denoise --patch-size 1024 -o Test_denoised/ Test/*.mrc,
-> There is an error TypeError: Cannot handle this data type: (1, 1, 5760), <f4

The movies are recorded on a Gatan K3 detector and the output from EMAN header is:

RO image file on unit 1 : Test/009.mrc Size= 4695571 K

Number of columns, rows, sections ..... 5760 4092 51
Map mode .............................. 2 (32-bit real)
Start cols, rows, sects, grid x,y,z ... 0 0 0 5760 4092 51
Pixel spacing (Angstroms).............. 10.25 10.25 10.25
Cell angles ........................... 90.000 90.000 90.000
Fast, medium, slow axes ............... X Y Z
Origin on x,y,z ....................... 0.000 0.000 0.000
Minimum density ....................... 0.0000
Maximum density ....................... 32767.
Mean density .......................... 2024.3
RMS deviation from mean................ 832.76
tilt angles (original,current) ........ 0.0 0.0 0.0 0.0 0.0 0.0
Space group,# extra bytes,idtype,lens . 0 0 0 0

---

It sounded like the correct input for topaz denoise would be MRC movie formats with > 1 subframe, so the algorithm can use both even and odd frames as separate groups in denoising. Is this correct? Could the issue be the Map mode of the movie?

Thanks,
Matt Larson

[Guillawme]

As I understand it, the input must be movies when you want to train a denoiser. But if you want to denoise micrographs using an already trained denoiser, then the input must be micrographs (so, the output of motion correction, with a single frame). I only have limited experience using topaz denoise, but it always gave good results, so I never needed to train a denoiser on my own movies.

Good luck!

[tbepler]

Exactly, topaz denoise expects micrographs as inputs. Even for training, the micrographs still need to be processed from the frames.