Raw File Optimization

What’s the best app to demosaic, denoise and sharpen your raw files?

NOTE: This article first appeared on the Luminous-Landscape.com. It is reproduced below in its entirety.

We are in the Golden Age of Photography, with sensors fine tuned for low and high ISOs producing pro quality images for printing and publication. I’m always pleasantly surprised—and, thinking back to my film days, a little shocked!—by the quality that we can extract from raw files. When combined with software algorithms for demosaicing, denoising and sharpening, well, as I said, it’s the Golden Age of Photography.

We shoot raw files to extract as much information from a scene as is technically possible. Photographers choose raw capture because they place a higher value on quality, legacy, and individual vision than on having ’ready-made’ machine JPEGs, compressed and sharpened. Over the decades I’ve been working in digital, I have yet to meet an image file that didn’t benefit from editing, and raw is the place to start. But are we getting the most from our raw files?

I’ve been using Lightroom for as long as it has existed, and Photoshop before that. Naturally, I’m curious—is Lightroom extracting all the data it can and optimizing it to provide the highest image quality possible from my sensor? I’ve spent a lot of time, effort and money to get that raw file, I want to ensure I’m getting the most bang for my buck. The only guarantee of that is through testing.

This is the first in a series of investigations examining how to extract the highest quality possible from a M43 20.1 megapixel OM-1 sensor. At 5184x3888pixels, the sensor is ideal for any and every use of photos right up to making two-page spreads in photo books and magazine and fine art prints as large as you need (see Finding the Sweet Spot in Photography).

Method

I’ve subjected the same six images to six different treatments. Each result is output to three commonly-used sizes: (1) for web use; (2) for 4K TV use and smaller prints; and (3) for larger prints and publication. The images were selected to push the sensor while representing different styles of my photography: landscapes shot during the day and evening; travel photography, and birds and wildlife.

Photos

Additional Comparisons:

I hadn’t intended to test higher ISOs as I rarely shoot above 3200. But, given the results from the ISO 3200 photos, I felt I should give high-ISO a shot, if only for a sense of completion. I  will discuss the results of these two shots separately, after discussing the other 4 together.

Raw Treatments

The five raw treatments are all commonly available, industry-standard apps for raw file demosaicing, denoising, and detail sharpening that are easily integrated into a Lightroom workflow. All processing was done on an M1, 16” MacBook Pro with 16GB of RAM.

• LrBase are raw files processed normally through Lightroom CC (v8.1). I edit only enough to breathe life back into the machine image—to re-create my experience and my vision in the field. This changes neither the original intent of the image nor its natural feel. I work with the pixels the camera captures, so there are no dropped-in backgrounds, skies or subjects. Adjustments are made as needed to Exposure, Contrast, White and Black Points, Highlights and Shadows as well as Colour and Tint. Additionally, I add Adjustment Masks to shape the light. Detail Sharpening is typically set to 60.

Each of the additional treatments began with the original raw file and took a reasonable 15 to 20 seconds to complete.

• LrEnhNR is the first of the five additional raw treatments. The LrBase file was run through Lightroom’s own Enhanced Noise Reduction algorithm set to 50 or 60; 75 for high ISO files. Finding the balance of noise removal while maintaining detail is critical. Afterwards, Detail Sharpening was typically set to 40, the Lr default for my ORF raw files.

Each of the following treatments began with the original ORF raw file. I did not use ‘File > Open In’ from within Lightroom as it creates unnecessarily large TIFFs.

• DxO: DxO Pure Raw (v4.7.0); DeepPRIME XD2s and DxO’s optical corrections were applied to the ORF. Output to DNG.
• ON1NN: ON1 No Noise AI (v2024.5) was applied to the ORF using the No Noise module set to ‘Standard’. The Tack Sharp AI module was not used. Micro Sharpening was set to the default 100 for all except for the Meru Forest file, set to 50; anything higher than 50 was too aggressive. Masking was not used. Output to DNG.
• TPZ: ORF raw file + Topaz PhotoAI (v3.4.3) Raw Denoise at Standard (Strong was too aggressive) and Sharpening set to ‘All’ for landscapes and ‘Subject’ for birds and wildlife. Some tweaking of the subject mask was needed; it was helpful to have that option and something only Topaz and ON1 have. Output to DNG.
• OMW: ORF was processed through OM Workspace (v2.3.3) with AI Noise Reduction plus the various built in lens corrections. Output as a TIFF; OMW does not output to DNG.

Output

After treatment, each DNG (or TIFF) was added back into Lightroom. The original LrBase edits were copied and pasted To each file, with some colour and distortion correction. JPEGs were output at 80% or 100% quality in sRGB colour space with Sharpening set to Screen. The three sizes of JPEGs represent common, everyday uses of photographs by the vast majority of photographers:

• 1500x1125px (1.7mp), at 80% quality, to represent the needs of social media platforms, blogs, forums and other web uses, plus for HD projection; e.g. at camera club presentations. Typically, social media sizes are smaller, so any differences at this larger size should be even more noticeable.
• 3840x2880px (11mp), at 80% quality, for 4K TV presentations, as well as laptop and desktop screens; e.g. wallpaper. This is also sized for prints a little larger than 9×12”, so it’s a test of prints up to that size as well.
• 4800x3600px (17mp), at 100% quality, is an ideal print size for photo competitions and to hang on your wall as it makes a 12×16” print at 300ppi, easily matted to 16×20” for framing or presentation. Note: each of the wildlife photographs had to be up-sized from the original to this size during the Lightroom export to JPEG. Normally, this would mean a hit on quality, but surprisingly . . . well, read on.

Results

The results I report are based on real-world uses of photographs, not on pixel-peeping. At times, I zoomed in to 200% to check sharpness, but viewing photos at this magnification is completely unrealistic. As photographers, it is all too easy to make judgements and pronouncements based on pixel-peeping simply because we can, rather than we need to— a real dis-service to the overall photograph. After all, it’s the emotional impact that attracts people to photographs, not the pixels. The details are helpful in creating that impact but what they look like at 200% is irrelevant. My feeling is, pixel-peeper internet bloggers are simply after clicks in the guise of truth, precision and service to photographers.

(1) Web- and projection-size (1500x1125px; 1.7mp)

Bottom Line: With correct sharpening and texture, any of the six treatments would produce fine web images. If projecting, then a run through LrEnhNR, DxO or Topaz would produce slightly sharper results.

(2) Screen Resolution and up to 9×12” (3840x2880px; 11mp)

When JPEGs were viewed at 100% on-screen, as would normally be the case for these photos, the differences became more apparent. All five raw treatments produced excellent results with a light edge going to LrEnhNR and DxO for their consistency across all photos, with smooth skies, balanced mountain and cloud detail, and micro-detail in foliage, fur, and feathers.

Below is a series of 100% crops from each of the 3840x2880px files arranged for side-by-side comparison. Viewing them like this, on screen, gives a sense of what they look like as web images (#1 above). Click on each image to open it to view it at full size.

To my eyes, the stand-outs continue to be  LrEnhNR, DxO, ON1 and Topaz. They are all equally sharp. Unfortunately, the ON1 and OMW files suffer from the same slight colour shift noted above.

By the way, this size of 3840x2880px is similar enough to the 3600x2700px required for a fine art 9×12” print. The results shown above, clearly  show that any of the five treatments (barring artefacts) would make excellent fine art prints at that size.

Not seen in the Topaz crop of Kili Blues is some colour mottling in the plain, blue sky area along with diagonal banding at the near pixel level. This showed up each of the multiple times I ran the file through. I have contacted Topaz about this and they are working on it. TopazLabs claim the ‘pin hole’ artefacts are ‘dead pixels’, yet they do not show on any other treatment.

 (3) Framed Print Resolution (4800x3600px or 17.3mp)

To best approximate viewing distance for actual prints, these larger files were evaluated on-screen at 50%. Again, judging at 100% is simply not realistic as only photographers and internet pixel-peepers, not buyers, view prints this close.

Despite the larger size (17mp vs 11mp), the results of these comparisons closely mirror the results of the previous files, but as expected, the fine differences start to reveal themselves.

Once again, the LrEnhNR and DxO versions are the best, but not by far. The ON1 and Topaz versions are equally good, but each suffers as they did previously from colour shift and artefacts respectively. I’d like to work out these problems as the sharpness and three-dimensionality are excellent.

High ISO Photos

The last two of the six photos I examined were shot at very high ISOs of 6400 and 12800. I now know that my previous scepticism is unwarranted and I’m proven wrong, again—the results are phenomenal, across the board. Web-sized JPEGs are virtually indistinguishable across the five treatments. At 4K size, the OMW file drops out of the running. Even upsized to 4800 pixels, the detail, colour balance and exposure are amazing and very printable amongst the LrEnhNR, DxO, ON1 and TPZ files. The LrEnhNR appears more naturally sharp with the others appearing more hyper-realistic, easily tamed by dialling down the Texture adjustment in Lightroom to –30 to –50. This is something I will discuss further in my conclusions.

Like the photos above, click to open each file to view them at full size.

One surprise was the quality of the LrBase photo of Junior at ISO 6400. I printed it to 4×6 as an ArtCard and you would never know it was shot at ISO 6400. This is something we as photographers easily lose sight of when pixel-peeping becomes the norm. Most prints on a lustre baryta or matte paper will not show the noise to the same extent as is shown at 100% on screen.

Conclusions & Discussion

I would love to say that one treatment stood out head-and-shoulders above the rest and was a clear winner, but I can’t. They all produce excellent results. There may be differences at the pixel level when viewed at 200%, but as I’ve made a point of saying, that 200% is an unrealistic yard stick to use.

Lightroom’s own Enhanced Noise Reduction is excellent throughout the ISO range, providing very natural-looking sharpness and micro-contrast while maintaining smoothness of skies and excellent overall three-dimensionality. DxO Pure Raw seems like the best of the raw optimizing apps, with Topaz and ON1 being equally good, but with a couple of artefacts that need more investigation. OM Workspace was disappointing in its ability to create clean, sharp images, particularly at higher ISOs.

Natural versus Hyper-realistic ‘Look’

One thing I kept noticing with all the photos is the very natural-looking, clean sharpness and presence of the LrEnhNR files, which contrasts with what I see as the hyper-realistic sharpness and smoothness of the DxO, ON1 and Topaz versions. The initial appeal of JPEGs from them is captivating, but to me they seem too real, the birds and lion looking more like museum specimens prepared for exhibit. They look great to the untrained eye, but to someone who knows nature, they appear, perhaps, too perfect. As a former ‘film guy’, maybe I’m just more tolerant of a more natural look, and less tolerant of the plasticky smoothness associated with these treatments. It’s like they are trying to emulate something that doesn’t really exist in nature.

My concern is that this is the way photography is going due to large sensors and this pixel-peeping drive towards minute, clean detail—detail and perfection at a level which one would never see in nature without the proverbial bird in the hand. Perhaps it’s the generative AI aspects of the algorithms, creating pixel-level detail that is not normally seen. Or, maybe it’s just me. Having been involved with nature interpretation for decades, the look comes across as a bit hyper-real, perhaps in an effort to make nature look glossy and catch attention, rather than showing nature as it is. But, as I said, this is only my perception, a feeling I have.

Final Assessment

Overall, you can’t go wrong with Lightroom Enhanced Noise Reduction, DxO Pure Raw, ON1 No Noise, and Topaz Photo AI Raw Denoise and Sharpening for demosaicing, denoising and sharpening raw files. If you’re looking for a clear winner, you won’t find one. They all perform brilliantly and, save for the Topaz and ON1 artefacts which are being looked into, any differences highlighted here are only noticeable upon direct comparison, which in itself, is unrealistic. Differences may be more or less noticeable depending on the raw files you start with and your personal workflow.

I realize these results may be different from what internet bloggers and vloggers have found, and I did not rank one above the other by unnecessarily splitting hairs—but I have no skin in the game either. I do not represent any of these companies, nor do I make a commission from links or profits from clicks. I am only reporting what I see and for the vast majority of photographs, any one of the apps, other than OM Workspace, does a superb job of cleaning up noise while maintaining and sharpening the fine details of foliage, fur and feathers, with the noted caveat of glossy perfection.

For me, I will continue to use Lightroom’s Enhanced NR and am building into my regular workflow both DxO and ON1. I will also be spending more time making some prints to see which versions look better on baryta and matte papers.

Up next in this series:

How well can Topaz PhotoAI and ON1 No Noise ’rescue’ images with motion blur? — now available!

---

Thanks for reading! If you have any questions about OM System, the quality it produces, or the photos and observations shown above, be sure to add a comment.

This work is copyright ©2025 Terry A. McDonald
and may not be reproduced in whole or in part without the written consent of the author.

Please SHARE this with other photographers or with your camera club, and consider signing up to receive an email notice of new blogs.

Have a look at my work by visiting www.luxBorealis.com and consider booking a presentation or workshop for your Nature or Photo Club or a personal Field&Screen workshop at Workshops.