Grainy raw images vs jpg / Expose noise reduction postprocessing options from libraw about rawpy HOT 11 CLOSED

Thanks for the reply! I will have a look. I am not entirely sure that noise is the correct term for what I am experiencing. When I open up the image in Image viewer on ubuntu it looks crisp and clear, but when I import it as a np array using rawpy it looks quite a lot different. would the program clean it up or am I doing something wrong?

from rawpy.

Cameras have quite advanced postprocessing algorithms that are tuned for a specific sensor etc. Reproducing that without help from the manufacturers is near impossible. I'm going to close this issue as there is nothing more that rawpy could do here. You could try to apply more image processing algorithms on top of the postprocessed image from rawpy. Sometimes simply downsampling helps with noise, if you can tolerate the loss in coordinate accuracy.

from rawpy.

Currently rawpy doesn't make all possible options from the underlying libraw library available, amongst them a few related to noise. Have a look at https://www.libraw.org/docs/API-datastruct-eng.html and search the page for "noise". If you make a list of the ones you need I can look at adding them for the next version. The ones currently available in rawpy are https://letmaik.github.io/rawpy/api/rawpy.Params.html#rawpy.Params but note the parameter names might be slightly different. There are none of the noise-related ones in there currently.

from rawpy.

I had a look myself and noticed that the libraw docs are slightly outdated and mention parameters that don't exist anymore. Since it was easy to do I added a few missing noise related ones, see the linked commit.
About your specific issue, what exactly do you open in Image Viewer? The jpg or the raw file? And similarly how do you use rawpy exactly? Do you look at the raw_image directly or do you look at the result of postprocess()?

from rawpy.

With image viewer, I am looking at the raw file (.cr2). When I import using rawpy I run postprocess() then use matplotlob.pyplot. Can you view the image without using postprocess()? This is when it is converted to RGB correct?

from rawpy.

No you can't really view an RGB image without postprocessing. There are many ways to postprocess, no idea what Image Viewer does and what libraries it uses to do that. Let me know if you want to experiment with the additional options I exposed, then I'll do an alpha release.

from rawpy.

I would be up for testing the added parameters!

from rawpy.

There you go: pip install rawpy==0.12.0a1

from rawpy.

How do I implement the 'fbdd_noise_reduction' ? . I've tried doing fbdd_noise_reduction = 0,12 etc and get
int' object has no attribute 'value'.

from rawpy.

You're right, I forgot to update the docs!
You would do fbdd_noise_reduction=rawpy.FBDDNoiseReductionMode.Full (or .Light or .Off (default)). See https://letmaik.github.io/rawpy/api/rawpy.Params.html for the other parameters.

from rawpy.

Hey,

It definitely produced better results! But still not as good as the native jpg's that the camera produces.

below are two different masks that I have produced, the first without the noise reduction, and the next with the noise reduction.

If you have any further suggestions as to what I could try I would be grateful, thanks for all the help so far!

from rawpy.