What is your screen resolution? Do you have some kind of scaling? I downloaded the same file and couldn't reproduce the problem, but then I noticed that the shadow in your screenshot is bigger than mine. I also cleared the cache and started with a fresh darktable config. When looking at the door on the right I can see differences again:ĭarktable export differences with dpr example imageĪgain, I tried different export formats. Since I was using ISO 1600 I tried the following raw file:Īgain, I used one denoise profiled with "wavelets auto" and a second denoise profiled with "non-local means" and a patch size of 7. I do notice very minor difference in the noise pattern between CPU vs GPU-enabled outputs, but that should not be a cause as the preview and export in darktable should both use CPU (or GPU). Also rename/delete ~/.cache/darktable to be on the safe side. I'm tempting to suggest that you rename ~/.config/darktable temporarily to have a fresh config just in case. Compare the history stack with mine, too, to see if something suspicious. The Geeqie viewer on the left looks the same as darktable at full zoom. In the screenshot below, I have two denoises, the first is default Wavelet Auto, and the second is default NLM with patch size set to 7. I downloaded an ISO 4000 RAW sample for the RX-100V, (not exactly your camera?), and tried to reproduce the problem but couldn't. I wonder what if you enable only the second instance of denoise? Trying to figure out if it's because of a second denoise (regardless of mode), or if it's because of the NLM mode, or a combination of having NLM as second instance. I used to have two denoise instances as well, and had some bugs, but haven't seen this.ĭo you have this same problem with all the RAW, or just some, or just this? If it happens to more than one, do you see a pattern (ISO, SS, aperture. Your finding of the second denoise confirms the differences in the screenshot in your first post, too. The fact that you narrowed it down to the second denoise indicates that there's nothing wrong with other modules/settings. However, I tried different formats, bit depth, profiles, intents, with and without high quality resampling etc. Here are my export settings in darktable (not lightroom): The exports start to look different from what I see in darktable when I apply the second denoise (profiled) module (having set mode to "non-local means"). I started with a new copy of the image, so I now have a clean history. The gamut check mode does not show anything, at least not in the mentioned white lines (see screenshot).I tried different export formats using different bit depths.I tried setting “high quality resampling” to “no”.I use AMaZE as demosaic, but also tried PPG and VNG.I export at the same size as the original RAW file.non-local-means), but again, I set the profile as above and increased the patch size to 7. One using chroma preset (wavelets auto) and manually set profile to “DSC-RX100M5A iso 1600”. I use two “denoise (profiled)” instances.So it does not seem to be related to the image viewer. When loading the exported JPG back in darktable I can see those noise/grain in darktable, too. Unfortunately I cannot share the whole image or raw file :/ (NOTE: The cyan colors are caused by the gamut check. You can see it in the screenshot I made, for example the structure of the white lines (on the left side you can see darktable and on the right the exported JPG). When I export that image as JPG (Quality 100), view it at 100% and compare the result to the image shown in darktable at 100% I can see that the exported JPG is not as soft as shown in darktable, but noisier/grainier. I used darktable to reduce noise of a high ISO image.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |