Benefits of Topaz denoise

Dear colleagues,
could anybody objectively comment, perhaps by just mentioning a personal experience, on the benefits of denoising micrographs with topaz, vs non-denoised? I really want to try it but with a very limited computational power. It feels like my pc will melt as soon as I start the denoising job.
Thank you in advance,


I tried using both original and denoised images to train topaz. In my case with a rod-shape 280 kDa protein, that most of my images have defocus -(1.2~2.8)um and CTF-fit 2.8~3.2 Å, I didn’t see denoised images helped, sometime even worse.



Are you asking whether denoising images is useful for visualization, or if denoising images is useful for subsequent Topaz picking? If it’s the former, you can make a small subset of images and denoised them. If it’s the latter, in our various tests we have not seen any particular advantage of using denoised images for Topaz picking.

I hope that helps.

Dear Feng, Alex, many thanks for this and sorry for not asking a clearer question. I was indeed thinking about Topaz picking, on denoised vs non-denoised micrographs. As I heard from a colleague at my university, denoised images did not help him as well, although Topaz picking has been great in general.
Thanks for sharing! Best wishes

1 Like

Does performing denoise on entire dataset with movies helps improve overall map reconstruction or density ? Is there any benefit with regard to the map except just for visualization?


@diffracteD We tried to look at that in this paper:

The algorithm is also noise2noise, the architecture is a little different and there is per-dataset training. IMO it may be possible that this approach can eventually help 3D classification, but software would have to account for the change in the shape of the signal.