Test Four NR/Sharpening Software Comparisons

[drr1531]

Recently I've been testing DxO2 and Topaz AI Sharpen. I've gone back and re-worked some images that were shot at high ISO in poor light. I compared results versus my typical PP for the past three or four years which involves basic lighting/color adjustments in LR and final lighting/color, NR, and sharpening handled in On1. My overall conclusions/observatons are nothing new:
- There is no one solution that fits all situations
- There are multiple ways to remove noise and sharpen which will yield essentially equal results
- The biggest difference in the various solutions is the amount of work, complexity, and time involved to yield desired results

In this post I'm going to compare results of four options for NR/sharpening: LR6 alone, LR+On1 Photo Raw, DxO PhotoLab2, and LR+Topaz AI Sharpen.

The image used for comparison below is posted at full resolution with no added output sharpening during jpeg conversion.

Top left: LR lighting, NR, and sharpening. Sharpen masking set to 92.
Top right: LR lighting, On1 NR and sharpening via USM.
Bottom left: DxO2 lighting, Prime NR set to auto, USM turned OFF
Bottome right: LR lighting, Topaz AI sharpen; sharpen module at default settings and grain OFF

D500, [email protected]
1/500s f5.6 ISO2500

Subscribe to see EXIF info for this image (if available)

Subscribe to see EXIF info for this image (if available)

Final comments based on all testing/experience to-date, not only relative to this particular image.

LR6 does and adequate job of NR/sharpening on images with lower levels of noise. It takes the most experience with the tools and may require a significant amount of masking/local adjustments. Sharpening tool is somewhat of a "black box" and just requires a lot of trial and error moving the sliders around. Hampered by lack of layers.

On1 Photo Raw is a pretty good supplemental layer editor to LR or standalone RAW converter. The lighting/color adjustment tools are very robust with some very useful presets/filters(e.g. pseudo HDR). The NR filter has basic adjustment sliders for luminance and color. Sharpening is very robust with options for "dynamic contrast"(aka micro contrast), "progressive sharpening"(black box), high pass, and USM. There are several masking tools that can yield excellent results but are fairly time consuming. Auto masking is poor and takes a LOT of processing power/time. Combining NR and sharpening with judicious masking can be very effective.

DxO2 works like magic ON SOME IMAGES based on content and depending on whether there is a camera/lens profile in the DxO database. Based on my limited testing the profiles seem to be where the magic lives in this software. Of all of the software I've used/tested it yields the best results for NR without destroying detail. Also the "lens sharpening" eliminates the need for capture sharpening and if the original is truly sharp no additional sharpening may be needed at all. DxO is a very good RAW processor with very robust tools/controls and "best in class" NR built in.

Topaz AI Sharpen is like magic.... on some images. There is a black box with very little user input/control. The "AI" aspect of the software analyzes the image and decides what details to sharpen and what constitutes noise to be supressed. There is a single slider for NR and one for sharpening. When it works it produces superior simultaneous NR/sharpening, even better than DxO in most cases(based on very limited testing). There are a couple of caveats. It is REAL slow. Even generating the on-screen preview takes several minutes for typical images. Also the Focus module can produce significant artifacts in the BG which requires masking/blending in a layer editor to correct.

 

[Butlerkid]

WITHOUT looking at the photo descriptions OR reading the test below the images, I picked #3 (bottom left) as the best. The others seem close but I had no problem settling on #3 as having better fine details.

I use DxO with only 3 things selected: NRPrime, CA, and distortion. However, the lack of detail and white spots on the forehead are not something I have ever seen when using DxO Prime. I believe if you turn OFF DxO Lighting, you will like the results a lot. (Yes, I realize you decided to use lighting on all 4 s/w programs, but IMHO it is distorting the results.)

 

[Pa]

WITHOUT looking at the photo descriptions OR reading the test below the images, I picked #3 (bottom left) as the best.

Same here, but on closer examination I think the one on lower right preserves a little more detail in the highlights. Perhaps a gentle adjustment of the highlight slider in DxO would fix that. The one on the lower right has a slightly different color cast and I have no way of knowing which is more accurate.

You have confirmed my decision to stick with DxO Photolab 2 as my do-it-all processor and photo editor. As Dan says, there are probably others that can do as well or better if you really know how to use them. But I've invested a year and a half in DxO and I'm getting pretty attached to it.

 

[Butlerkid]

Same here, but on closer examination I think the one on lower right preserves a little more detail in the highlights. Perhaps a gentle adjustment of the highlight slider in DxO would fix that. The one on the lower right has a slightly different color cast and I have no way of knowing which is more accurate.

You have confirmed my decision to stick with DxO Photolab 2 as my do-it-all processor and photo editor. As Dan says, there are probably others that can do as well or better if you really know how to use them. But I've invested a year and a half in DxO and I'm getting pretty attached to it.

But the detail would be in the highlights if DxO Lighting were turned off.......

 

[Pa]

But the detail would be in the highlights if DxO Lighting were turned off.......

What do you mean by "DxO Lighting"? Do you mean "Smart Lighting"? I seldom use that.

 

[Butlerkid]

What do you mean by "DxO Lighting"? Do you mean "Smart Lighting"? I seldom use that.

Either Smart Lighting or Clear View. I don't know which was used. I don't use either one in DxO as I think they are often very heavy handed and I get better results simply setting black and white points and adjusting highlights and shadows as needed back in ACR. That's the reason I say lighting is distorting the results in DxO and maybe the other 3 s/w packages as well.

 

[Pa]

Either Smart Lighting or Clear View. I don't know which was used. I don't use either one ...

Smart Lighting and Clear View are very different. I don't like Smart Lighting but I often use Clear View. I agree that it can be heavy-handed, but that's what the slider is for.

That's the reason I say lighting is distorting the results in DxO and maybe the other 3 s/w packages as well.

I agree.

 

[drr1531]

You guys are overthinking this. Karen, yes by definition the lighting is different in the DxO conversion vs the LR conversion in the other three. I think I left smart lighting turned on in DxO but that would supress highlights if anything. I probably supressed them more in LR. But forget about the lighting on the critter's head. I made no attempt to optimize lighting on any of them. Just pulled things into the ballpark. I'm not saying lighting isn't important in pulling detail out on fur/feathers. Just saying it's not necessary to optimize it to compare the NR/sharpening results.

For comparing the NR and sharpening IMO the eyes, detail in the nose, sharpness of the whiskers, and noise level on water and shadows tell the tale. IMO the detail is slightly better in the AI Sharpen image but the DxO is so close as to be a moot discussion. If using either one or the other software there's certainly not enough difference to warrant a change. At least not based on these results. How you get there is a different story. Even though on this particular image the AI Sharpen is slightly better it's not enough to warrant living with the crazy processing time that it takes. I'm convinced enough to spring for the cost of DxO.

In fairness to LR, I tweaked the sharpening mask until pretty much nothing but whiskers and other bright edges were sharpened. That's why the nose looks so smudgy. Easily corrected with a few brush strokes. Not that it would matter in the overall result.