Issue: Pattern Caused by Unsharp Mask and Perspective Correction of High ISO Photos

Maybe later. It would be more interesting to see if you can produce the effect with one of your own files. The effect can best be seen in dark, uninteresting areas. The pattern might depend on other settings too (you can see the tools I used in the screen capture) and is not always the same.

Tried other settings that produced slightly less pronounced effects. The following should suffice, but other settings could accentuate or attenuate the effect, change the pattern etc.

The following preset should do the trick.
Pattern Preset.preset (10.4 KB)

I was thinking to try your file, because you have different camera than mine. And if you have the files you are testing, it would save me the time to have to shoot for this with a different camera. I wanted to try on your file, I have few theories what is happening, but I would have to test the settings for the file. If you can, upload the RAW file here, later.

it is a usual result of crude image correction ( perspective ) applied by DxO - very easy to reproduce… USM just makes banding more visible ( just like with ACR / LR where crude optics correction induce noise banding )

https://postimg.cc/HrDPp9Pb

and just to show that this is not just a on-screen rendering in DxO - output and open in PS

save the image below ( link ) on your computer and view @ 100% - do not view directly in browser from the link

https://postimg.cc/8J0txR6J

I have never seen any problem of this type with perspective correction activated on high iso files.
But for all that, it would never have occurred to me to combine lens sharpness correction AND USM correction.
The result of this combination (which I also see with these same settings) seems normal to me.

Other DXO Photolab optical corrections (Vignette and Distortion correction) also causes banding, see:

Lex

Unsharp Mask is basically edge detection + contrast type filter. It is the oldest form of sharpening. Find edges and increase the contrast of pixels. Make white more bright and dark pixels darker. Giving an illusion of sharpening.

The purpose of the Unsharp Mask tool is to “sharpen” an image which done by this crude way will exaggerae the edge contrast of what is already there. Including artifacts. The tool makes a blurred copy of the original picture, then subtracts the original from the blurred copy, leaving the finest details, which can then be enhanced.

Intensity: sets the amount of sharpening to be applied to the whole image. Basically intensity slider is for how much or how strong effect you want.

Radius: sets the thickness of the edges to be sharpened.

Threshold: sets the level above which details will be sharpened, and below which they will be left as they are, making it possible to avoid sharpening the smallest details that look just like noise.

Edge offset lets you homogenize the sharpness between the center and the edges of an image.

If you are seeing really bad pattern or artifacts its because you are exaggerating what is there. Its not rocket science. It you don’t want to see the imperfections don’t go to far with unsharp mask effect.

If there is a specific bug or issue with particular camera sensor, I would love to test it with the image that shows that. Otherwise to me this looks like wrong expectations about how the tool should perform.

1 Like

chatgpt as usual

No name comment, as per usual. Dude, for noname guy you often sound like a grungy chat bot. Try human for a change. You can start with a name. What is your name?

1 Like

as usual your posts 99% are copy-paste from somewhere else … good bye.

As usual you are acting like a grumpy bot. But hey, what else to expect from some no name grumpy bot. And for the record, you lame assumption is wrong, per usual. If you insist of trolling, try more finesse, dude.

no assumptions - plain facts - captain obvious is at his best in being didactic ( as if somebody here needs him to repeat what USM is ) :

Unsharp Mask is basically edge detection + contrast type filter. It is the oldest form of sharpening. Find edges and increase the contrast of pixels. Make white more bright and dark pixels darker. Giving an illusion of sharpening.

The purpose of the Unsharp Mask tool is to “sharpen” an image which done by this crude way will exaggerae the edge contrast of what is already there. Including artifacts. The tool makes a blurred copy of the original picture, then subtracts the original from the blurred copy, leaving the finest details, which can then be enhanced.

Intensity: sets the amount of sharpening to be applied to the whole image. Basically intensity slider is for how much or how strong effect you want.

Radius: sets the thickness of the edges to be sharpened.

Threshold: sets the level above which details will be sharpened, and below which they will be left as they are, making it possible to avoid sharpening the smallest details that look just like noise.

Edge offset lets you homogenize the sharpness between the center and the edges of an image.

Maybe if you get the foot our of your mouth, you can enlighten us with noname wisdom than, instead of being obnoxious troll hiding behind empty name.

You forget to quote a passage from the user guide, which passage has a certain importance:

Using the Unsharp Mask
The Unsharp Mask correction is disabled by default. It is unnecessary for JPEG files, as in-camera processing has already sharpened them, and it is usually unnecessary for RAW images for which a DxO Module is available. This means its use is really confined to unsharpened JPEG files and RAW files without a DxO Optics Module. In the latter instance, we advise fine-tuning the Unsharp Mask settings, and then creating a preset.

We recommend that you try fine-tuning the three sliders using these starting values: Intensity = 100, Radius = 0.5, and Threshold = 4. For most images, Threshold should stay within a range from 4 to 10. Radius determines how subtle the correction is: excessive values will result in halos. Finally, you can set the Intensity slider up to 200.You forget to quote a passage from the user guide, which passage has a certain importance:

2 Likes

Yeah. DXO considers their lens sharpness to be better way to sharpen images for the RAW files supported by their optics module and its true. Lens sharpness is more tailored to particular lens, and it does no apply same sharpening level to whole of image, but based on measurements from the DXO for a particular lens.

Unsharp mask is there I think to help when there is no lens sharpening option available, like it is in some situations, and when it comes to JPEGs that have not been processed by other sharpening effects .

For more control they offer four sliders including one that can simulate in a crude way what lens sharpening does. Edge offset lets user apply the sharpness between the center and the edges of an image.

Adding unsharp mask on top of lens sharpening and doing it to the extreme will only exaturate any imperfections and artifacts in the image and generally is not a good practice unless someone is trying something creative with it.

As I’ve said before Unsharp Mask basically detects the edges and adds contract to the edges, creating an illusion of sharpening, with often nasty halo effect and or sharpening artifacts. Its a tool that has been around forever and its very blunt tool in most cases.

Adobe has a good article on how Unsharp Mask works.

Sharpen photos with the Unsharp Mask.

“The Unsharp Mask is designed to enhance the details in an image,” says Photoshop expert Jesús Ramirez. “Photoshop is not creating the details, it’s just creating the illusion that there is more detail. And it is doing so by creating contrast on the image.”

The Unsharp Mask increases the image contrast along the edges of objects in a photo. The effect doesn’t actually detect edges, but it can identify pixel values that differ from their neighboring pixels by a certain amount. From there, the mask will increase the contrast of the neighboring pixels, making the light pixels lighter and dark pixels darker. This increased contrast makes objects more identifiable and creates an illusion of more detail.

Keep in mind, if you overshoot the sharpening, it creates a halo effect around the edges of objects. The amount you sharpen a photo is up to your personal preference and artistic vision, but you can mitigate over-sharpening halos by adjusting sliders and experimenting with the Unsharp Mask settings.

https://www.adobe.com/creativecloud/photography/discover/unsharp-masking.html

Over the years there have been many new tools added to avoid the pitfalls of unsharp mask like halos at the edges which are distracting , unfaltering and noticeable. Some tools like Photoshop Smart Sharpen allows to adjust where sharpening is applied, light or dark edges and by what amount.

But its still very limiting overall. There are some AI solutions like Topaz sharpen AI and clever algorithms for Photoshop that help to apply different look to the image. One I like is: NBP Ultrasharp for Adobe Photoshop but I don’t know if its still available. It creates a look that I’ve not seen any other program do. Its quite unique.

Overall when working with RAW files, I find Lens sharpness DXO offers to be great tool for sharp looking images that are supported. Usually its all I use. But one can go overboard with that one just as one can do too much with each of these tools.

Thank you all for your inputs…but I’m pointing at something different that DxO could possibly consider. Let’s start with a small detour.

Human visual perception resolves differences of about 1/250, which is why jpeg can do with 8 bits/pixel. Now, for something to become visible, differences need to be amplified in such ways as to make them bigger than 1/250. If images were processed with 16 bit precision, an amplification of about 250 would just do that. If images were processed with 32 bit precision, the amplification needed would be around 1.6 Million…and I suppose that processing with perspective and USM does not produce such effects.

→ My point is, that, if images were processed with higher precision, the effects described in the OP could be avoided.

I’m not sure what I see on your example.
Maybe I’m wrong, but it seems to me that I see unsharp mask applied on a very noisy image.
In this case, unsharp mask should accentuate noise. And it seems to me this is what I see : sharpened noise. And this is not what you want.

Did you try to reduce noise (and get a no noisy image) before applying unsharp mask ?

When sharpening image like this (I mean a noisy image with a sharp subject above a blurred background), with this kind of sharpener (unsharp mask), I always try to get a fully denoised background before applying unsharp mask (the blurred background is the best place in your image to see if your image is fully denoised).

Here deepprime at 10 % seems to be not enough to fully denoise your image.

And is it really different without perspective correction ?

It’s not about the image. I took it at high ISO with focus is slightly off etc. I did all this to test what it takes to derail the app used to profile that photo.

Pushing USM to the extreme made the image look less ugly than I had expected, but with higher USM values came the more distinct, moire like patterns. Because the pattern moves and scales with the image, it’s not moire between image and screen pixels (moire needs superimposed patterns that are slightly different) which, at the moment, leaves bit depth, or the lack of enough bit depth, as a possible cause for the issue.

I was talking about the noname image.
I saw patterns on yours, but noname image does not seem to really show the same thing.

But anyway sharpening a noisy and slightly out focus image is maybe not a fair test.

Does this happen on tack sharp and fully denoised image ?

This is good to know. So Unsharp Mask is really legacy for images without a DXO module available for their camera/lens combination. I find I never use Unsharp Mask in DXO, and now I know why.

1 Like