Excessive Sharpening

[Reimar]

Well, after 10 years and more than 10,000 images, it finally happened... a QC failure.

What bugs me is that I don't see it.  I have been doing some minimal sharpening at the same setting for all these images.  I don't see any hint of artifacts in the image that failed. 

It's a landscape shot with a very sharp center, but edges are considerably softer.  It's a property of my Nikon 24-70 lens at f8, the D810, and the receding beach scene.

Is there a tribunal I can appeal this too?

I guess I'll have to keep the center part of this image softer (for no legitimate reason).

 

This is the image:

 

 

Here is a 100% crop of the center

 

 

And here is the distant edge:

 

I see the above images are downsampled.  To see the real thing I have to give links to my gallery:

 

http://www.pbase.com/reimar/image/163055195

http://www.pbase.com/reimar/image/163055196

Just a note for those of you with high-rez monitors (mine is 4K).  I usually have the internet screen set to 125% to better see text.  To evaluate these images you need to set this back to 100% for a moment.

[Guest]

There's certainly pixelation/artifacts in the sharpened centre but that could be a consequence of the hosting of the jpeg for viewing here - The woman standing by the breaking wave, as an example, has what look like jaggies on her arm...which doesnt make sense but it just looks wrong.

 

Would be nice to see in photoshop to evaluate properly. Not sure what you mean by the 100% since I see it at the size you post......

[Reimar]

Yeah, no pixelation is present.  To see the linked images like you were looking at it in photoshop at 100%, there can't be any zoom on your internet software.  For example, I use the Firefox "Theme and Font Changer" to set my zoom at 125% to see internet content at a size I can see.  I need to dial this back to 100% (or no) zoom to see the linked images as they appear in photoshop.

I know; technology is a pain sometimes.

[Guest]

Yeah, no pixelation is present.  To see the linked images like you were looking at it in photoshop at 100%, there can't be any zoom on your internet software.  For example, I use the Firefox "Theme and Font Changer" to set my zoom at 125% to see internet content at a size I can see.  I need to dial this back to 100% (or no) zoom to see the linked images as they appear in photoshop.

I know; technology is a pain sometimes.

 

Well, I've just looked at the originals on pbase and we will have to disagree - I'm seeing plenty of sharpening artifacts like a pixelation. Very obvious in the child looking out to sea alongside the clild with the green ball - his face looks like lego on one side..... it's occuring in many contrast edges.

[Alamy]

Well, after 10 years and more than 10,000 images, it finally happened... a QC failure.

What bugs me is that I don't see it.  I have been doing some minimal sharpening at the same setting for all these images.  I don't see any hint of artifacts in the image that failed. 

It's a landscape shot with a very sharp center, but edges are considerably softer.  It's a property of my Nikon 24-70 lens at f8, the D810, and the receding beach scene.

Is there a tribunal I can appeal this too?

I guess I'll have to keep the center part of this image softer (for no legitimate reason).

 

This is the image:

 

 

Here is a 100% crop of the center

 

 

And here is the distant edge:

 

I see the above images are downsampled.  To see the real thing I have to give links to my gallery:

 

http://www.pbase.com/reimar/image/163055195

http://www.pbase.com/reimar/image/163055196

Just a note for those of you with high-rez monitors (mine is 4K).  I usually have the internet screen set to 125% to better see text.  To evaluate these images you need to set this back to 100% for a moment.

 

Just to reiterate and remove confusion - the crops here are not showing at 100% - when viewed at 100% ("Original" size on your pbase links) the sharpening artifacts are very clear across the image.

 

Alamy

[Reimar]

Hmm.  We're obviously seeing different displays of these images.  I've not seen pixellation as an artifact of sharpening.

I am curious though.  I've PMed you my e-mail address.  If you respond I'll send you the whole jpg to look at in photoshop.

[Reimar]

Thanks for the response Alamy.  I know you're not seeing halo's because I use minimal sharpening and I'm very careful about that.

What do these artifacts look like on your screen? 

It would help me identify what to avoid in future, because I'm only seeing a very clear image in the center.

[Alamy]

Thanks for the response Alamy.  I know you're not seeing halo's because I use minimal sharpening and I'm very careful about that.

What do these artifacts look like on your screen? 

It would help me identify what to avoid in future, because I'm only seeing a very clear image in the center.

 

Lots of jaggies - this isn't a haloing issue. The edge of the face as Geoff points out is a good place to see this but also look along the edge of the orange rubber ring, the wing of the large bird in the middle - steps/jagged edges. Usually typical of over-sharpening - if this hasn't been sharpened then something else may have caused the issue. We also used to see issues like this in previous years when images were excessively upsized - not saying that's the case here but it's a similar effect.

 

Cheers

 

Alamy

[hdh]

The boys face next to the green ball is really very obvious. 

He looks a bit like a freak. 

[Reimar]

OK, I do see jaggies everywhere at or above 200% in photoshop.  But that's the case in all images, with or without sharpening.  I'm not sure why I can't see that at 100% in this image, when others can.

I'll take off any sharpening and see if that goes through QC.

[Ed Rooney]

The amazing and important thing here, Reimar, is that you have submitted 10,000 images without a QC fail. It shows that you have a great workflow and strong mental focus. Your amazing record is surely why Alamy were nice enough to address the issue in the forum. Good luck with the next 10 K . . . and please don't send a 16x20 print to Alamy.  

[IDP]

OK, I do see jaggies everywhere at or above 200% in photoshop.  But that's the case in all images, with or without sharpening.  I'm not sure why I can't see that at 100% in this image, when others can.

I'll take off any sharpening and see if that goes through QC.

The reason is that you've said you're using a 4k monitor. Because of the pixel density, you need to view at 200% to get the equivalent of 100% on a standard monitor.

 

Hope this helps

 

Ian D

[spacecadet]

I couldn't stop my browser from resizing, but when I enlarged manually it was clear.  It's not visible at about 5000 long side. Another reason I downsize everything to 4000.

Incidentally how long did yo have to wait for your fail?

[Niels Quist]

Great thread - thank you for sharing - and Reimar, though it may hurt your professionalism at the moment, get over it, it doesn't mean a thing unless you continue not taking notice. Could even be seen as part of your professionalism.

 

Niels

[Reimar]

 

OK, I do see jaggies everywhere at or above 200% in photoshop.  But that's the case in all images, with or without sharpening.  I'm not sure why I can't see that at 100% in this image, when others can.

I'll take off any sharpening and see if that goes through QC.

The reason is that you've said you're using a 4k monitor. Because of the pixel density, you need to view at 200% to get the equivalent of 100% on a standard monitor.

 

Hope this helps

 

Ian D

 

That would be a frightening prospect for me.  I think images look horrid at 200%.

If there were underlying jaggies, wouldn't the higher res screen show them up even more?

I must say, that 32" 4K screen has been a mixed blessing.

[Reimar]

I couldn't stop my browser from resizing, but when I enlarged manually it was clear.  It's not visible at about 5000 long side. Another reason I downsize everything to 4000.

Incidentally how long did yo have to wait for your fail?

A day.  We'll see how long the re-submit takes.

[John Mitchell]

This is probably no help at all, but about three years ago I had an image fail QC due to JPEG (compression) artifacts. They can look like "jaggies" apparently. Might that be the problem?

[hdh]

It could indeed be the 4K Monitor, because the pixels are so small that everything looks sharp. 

I recently started a thread about 4K Monitors, to find out about others experiences.

 

I have one ordered a large 32" Monitor and hope I can still do QA work on it, despite the smaller pixels. 

If not, I have a dual monitor setup with my old HD Monitor@1920x1200 as second screen, which I could revert to for QA. 

[Reimar]

This is probably no help at all, but about three years ago I had an image fail QC due to JPEG (compression) artifacts. They can look like "jaggies" apparently. Might that be the problem?

I can't be sure since I'm not seeing jaggies at 100%.  But somehow I doubt it.  In my experience compression artifacts tend to be posterization and larger blocks rather than pixel level jaggies that others are describing.

Plus, I sent Geoff a jpg to check in photoshop.  The Alamy image was jpg 11, the one on my gallery was jpg 10 and the one to Geoff jpg 8.  Geoff said the jpg 8 in photoshop was better (less obvious jaggies) but still noticeable.  So I don't think it's compression.

 

I'm going to do some testing to see if I can see an effect.  Like Ian said, I'll need to check things at 200% since I don't see issues at 100%.

I've been using Nikon's "clarity" slider since the D810 came out, so I'll test that.  Also I just switched from Nikon Capture NX2 to the Nik Viveza 2 plug-in for PS CC 2015 and using the new "structure" slider.  My sharpening algorithm has been the same for many years.  I may throw up some samples if it looks interesting.

[Duncan_Andison]

 

OK, I do see jaggies everywhere at or above 200% in photoshop.  But that's the case in all images, with or without sharpening.  I'm not sure why I can't see that at 100% in this image, when others can.

I'll take off any sharpening and see if that goes through QC.

The reason is that you've said you're using a 4k monitor. Because of the pixel density, you need to view at 200% to get the equivalent of 100% on a standard monitor.

 

Hope this helps

 

Ian D

 

 

+1. Very true. It also applies to Retina screens. I use a 2560x1440 at 27inch and that is the max resolution I'd use at that size. Anything more and the pixel density increases and provides false sharpness.

 

 

 

OK, I do see jaggies everywhere at or above 200% in photoshop.  But that's the case in all images, with or without sharpening.  I'm not sure why I can't see that at 100% in this image, when others can.

I'll take off any sharpening and see if that goes through QC.

The reason is that you've said you're using a 4k monitor. Because of the pixel density, you need to view at 200% to get the equivalent of 100% on a standard monitor.

 

Hope this helps

 

Ian D

 

That would be a frightening prospect for me.  I think images look horrid at 200%.

If there were underlying jaggies, wouldn't the higher res screen show them up even more?

I must say, that 32" 4K screen has been a mixed blessing.

 

 

4k and retina screens are great to work with but not to QC with. I would stick to 2560x1400 resolution up to 27inch. Anything more and you will miss SoLD shots as well as over sharpening. Retina screens also high density pixels screens and cannot be trusted for QC work.

[wiskerke]

A good test usually is to apply your final sharpening twice.

If it falls apart, the initial setting should be halved.

 

Traditionally one should not sharpen stock because clients would know how to use output sharpening.

 

wim

 

edit: would have known (?)

[Brian Yarvin]

I'm noticing here that I can't give a green up arrow to Alamy for coming on  the board and explaining their position. I really appreciate this. Knowing that there are knowledgeable people behind those QC judgments is quite assuring.

 

Alamy responder, you deserve an up arrow!

[dustydingo]

 

This is probably no help at all, but about three years ago I had an image fail QC due to JPEG (compression) artifacts. They can look like "jaggies" apparently. Might that be the problem?

I can't be sure since I'm not seeing jaggies at 100%.  But somehow I doubt it.  In my experience compression artifacts tend to be posterization and larger blocks rather than pixel level jaggies that others are describing.

Plus, I sent Geoff a jpg to check in photoshop.  The Alamy image was jpg 11, the one on my gallery was jpg 10 and the one to Geoff jpg 8.  Geoff said the jpg 8 in photoshop was better (less obvious jaggies) but still noticeable.  So I don't think it's compression.

 

I'm going to do some testing to see if I can see an effect.  Like Ian said, I'll need to check things at 200% since I don't see issues at 100%.

I've been using Nikon's "clarity" slider since the D810 came out, so I'll test that.  Also I just switched from Nikon Capture NX2 to the Nik Viveza 2 plug-in for PS CC 2015 and using the new "structure" slider.  My sharpening algorithm has been the same for many years.  I may throw up some samples if it looks interesting.

 

 

I steadfastly stay away from the bells and whistles of CC like Structure and Clarity--I experiemented with some of them ages ago and it didn't take long before I realised they, on my tests, caused more QC-type problems than they solved . . . again, just my experience. But one thing I do remember reading somewhere quite recently was that 'Structure" applies a particular sharpening parameter . . . added to Clarity and any other Sharpening, I shudder to think of the accumulative effect.

 

Ad you say, you are going to look at those very adjustments, so notwithstanding the monitor issue, I suspect you may find some room for "improvement" there, especially if you have only recently added the use of the Structure adjustment to your workflow.

 

good luck.

 

dd

[M.Chapman]

Well, after 10 years and more than 10,000 images, it finally happened... a QC failure.

What bugs me is that I don't see it.  I have been doing some minimal sharpening at the same setting for all these images.  I don't see any hint of artifacts in the image that failed. 

It's a landscape shot with a very sharp center, but edges are considerably softer.  It's a property of my Nikon 24-70 lens at f8, the D810, and the receding beach scene.

Is there a tribunal I can appeal this too?

I guess I'll have to keep the center part of this image softer (for no legitimate reason).

 

I see the above images are downsampled.  To see the real thing I have to give links to my gallery:

 

http://www.pbase.com/reimar/image/163055195

http://www.pbase.com/reimar/image/163055196

Just a note for those of you with high-rez monitors (mine is 4K).  I usually have the internet screen set to 125% to better see text.  To evaluate these images you need to set this back to 100% for a moment.

Reimar,

 

Excellent. Many thanks for posting links to 100% versions. Must admit my initial reaction was that they looked good, but after reading the posts and a closer look I have to agree with QC. Makes me realise how reliant I am on maintaining a consistent workflow, I think my eyesight can't be as good as it was when I setup my workflow. I need to be more careful in future.

 

Looking really closely (200%) I note there are numerous short horizontal rows of pixels that oscillate between lighter and darker shades. There are lots in the water near the shore on the left and some on the patio just under the dark blue parasol nearest the centre, here they appear as short vertical bars. I'm not sure what's caused those. JPG compression? Moire? Demosaicing? Once I'd spotted them I can still see them at 100%

[Guest]

 

This is probably no help at all, but about three years ago I had an image fail QC due to JPEG (compression) artifacts. They can look like "jaggies" apparently. Might that be the problem?

I can't be sure since I'm not seeing jaggies at 100%.  But somehow I doubt it.  In my experience compression artifacts tend to be posterization and larger blocks rather than pixel level jaggies that others are describing.

Plus, I sent Geoff a jpg to check in photoshop.  The Alamy image was jpg 11, the one on my gallery was jpg 10 and the one to Geoff jpg 8.  Geoff said the jpg 8 in photoshop was better (less obvious jaggies) but still noticeable.  So I don't think it's compression.

 

I'm going to do some testing to see if I can see an effect.  Like Ian said, I'll need to check things at 200% since I don't see issues at 100%.

I've been using Nikon's "clarity" slider since the D810 came out, so I'll test that.  Also I just switched from Nikon Capture NX2 to the Nik Viveza 2 plug-in for PS CC 2015 and using the new "structure" slider.  My sharpening algorithm has been the same for many years.  I may throw up some samples if it looks interesting.

 

 

Bit late getting back to this. I would suspect that it's clarity doing the damage here, it's not a compression issue. The problems look like they are occuring at contrast borders which is where sharpening is most obvious. As I mentioned to Reimar, if I have a slightly less sharp file at 60MB, I just downsize to 50MB or so with bicubic sharper which does the trick. Given the sheer size of a 36mp file, downsizing to 50MB would act as enough of a sharpen (I would guess) to overcome most lens issues.

 

And to echo Edo's post, top marks to the Op for posting this...... I now know for certain that 36mp is the devil