Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: rabanito on March 14, 2019, 03:39:01 pm
-
I made three copies of this image and sharpened one with NIK Output and a second one with Photokit using their defaults for this size (800x540) and for the web
Both sharpening tools have (for my taste) oversharpened, one more than the other.
Is there a reason for that?
My experience with sharpening for Inkjet has been always satisfactory but not for the monitor display.
I know, one can adjust it but perhaps I am missing something?
-
This subject matter definitely does not need sharpening. Modern cameras already produce images way sharper than in the film era.
Lightroom has output sharpening and enough choices (none, low, medium, high) to satisfy everyone. When I post to Flickr, I sometimes choose "none" or "low," as Flickr's algorithm tends to oversharpen by itself.
-
I made three copies of this image and sharpened one with NIK Output and a second one with Photokit using their defaults for this size (800x540) and for the web
Both sharpening tools have (for my taste) oversharpened, one more than the other.
Is there a reason for that?
Hi,
In my opinion, the downsampled image (original.jpg) has accumulated a lot of downsampling artifacts (halos and aliasing artifacts).
Those get only exaggerated further by sharpening them. Best is to try another method before sharpening.
Cheers,
Bart
-
I am speaking of sharpening for the screen and of course I believe that in this case it is WYSIWYG. I can adjust sliders until I am satisfied and that's it.
When sharpening for print I have no way of judging how much sharpening is needed. It could be that I'd need some trial and error :o.
Interestingly, IMO the sharpening for Inkjet printing is always good enough "as is", with the default settings for the given print.
Instead, the offered settings from the same tools tend to oversharpen for my taste when preparing the image for the screen.
So I wondered why. I understand what Bart says about downsampling but this (downsampling) is something that occurs all the time and in my case I always use Automatic Resampling in Photoshop.
I expected that if a tool doesn't oversharpen for Inkjet Printing it would perform similarly well when sharpening for the screen.
It is not of vital importance, just to increase my general knowledge. That's why I originally put it in the Coffee Corner, Jeremy
But it is OK too ;D
-
I am speaking of sharpening for the screen and of course I believe that in this case it is WYSIWYG.
On your display, at 100% yes. On another display, maybe not.
-
On my monitor it seems the original is not sharp... (?)- is it- i mean the image itself.
If you sharpen it it becomes ugly.
-
I understand what Bart says about downsampling but this (downsampling) is something that occurs all the time and in my case I always use Automatic Resampling in Photoshop.
I don't know what default downsampling method your Photoshop is set to, but Photoshop used to be very bad for resampling (using the 'Sharper' method, which is horrible). Even regular Bicubic is not very good, unless one first applies some blur to avoid aliasing.
Cheers,
Bart
-
I don't know what default downsampling method your Photoshop is set to, but Photoshop used to be very bad for resampling (using the 'Sharper' method, which is horrible). Even regular Bicubic is not very good, unless one first applies some blur to avoid aliasing.
Cheers,
Bart
"Bicubic Automatic"
But I must admit that till you mentioned it I hadn't given that a thought. It's over my head :)
-
On my monitor it seems the original is not sharp... (?)- is it- i mean the image itself.
If you sharpen it it becomes ugly.
ISO 3200 f/1.4
The nose is sharp. But I agree, I didn't choose a good example
-
On your display, at 100% yes. On another display, maybe not.
I understand - since it depends on the display - that it is irrelevant whether I sharpen for the web or not, provided that the original image is sharp.
-
"Bicubic Automatic"
But I must admit that till you mentioned it I hadn't given that a thought. It's over my head :)
That setting is just fine. It will automatically select the Bicubic algorithm based on if you're sizing up or down. Nothing wrong with how PS does this! Users of course can hose any image by over sharpening when they shouldn't.
On my screen (a NEC PA271Q), the original looks a tad soft.
One major issue with sharpening anything visually is that as you move some sharpening controls up, feel it's a tad too much, then lower, the effect appears to look 'too soft' in comparison. It's one reason why people over sharpen. The eye is playing ticks on them. You need to walk away or refresh your view, then reexamine the image, of course at 1:1 (100%).
-
That setting is just fine. It will automatically select the Bicubic algorithm based on if you're sizing up or down. Nothing wrong with how PS does this!
There is a lot wrong with how PS does this! (see attached) I have my CS6 preferences set to regular Bicubic, which produces slightly fewer artifacts on downsampling, and I use Smoother for upsampling by manual override, but that's not optimal either.
Just look at the ringing artifacts that it caused around the cat's whiskers, to point out just one of the many obvious flaws. Not fine at all.
Cheers,
Bart
-
There is a lot wrong with how PS does this! (see attached)
Maybe with an ancient version of PS you use. ;D
One can control this with modern versions of the product too:
https://www.creativebloq.com/adobe/photoshop-cc-image-sizing-61411983 (https://www.creativebloq.com/adobe/photoshop-cc-image-sizing-61411983)
I suspect millions of images have been resampled with these algorithm, few singing such an alarm. Sounds like FUD.
For the OP:
https://petapixel.com/2010/02/17/use-bicubic-sharper-for-web-resizing/ (https://petapixel.com/2010/02/17/use-bicubic-sharper-for-web-resizing/)
-
Maybe with an ancient version of PS you use. ;D
One can control this with modern versions of the product too:
https://www.creativebloq.com/adobe/photoshop-cc-image-sizing-61411983 (https://www.creativebloq.com/adobe/photoshop-cc-image-sizing-61411983)
I suspect millions of images have been resampled with these algorithm, few singing such an alarm. Sounds like FUD.
For the OP:
https://petapixel.com/2010/02/17/use-bicubic-sharper-for-web-resizing/ (https://petapixel.com/2010/02/17/use-bicubic-sharper-for-web-resizing/)
My webpage from 2004 on the subject of downsampling, https://www.bvdwolf.nl/foto/resample/down_sample.html , is quoted often as an illustration and warning against using poor downsampling algorithms, like Photoshop's Bicubic Sharper.
Many threads here on LuLa have discussed the much better algorithms that are available, including the ones in Lightroom. But Photoshop's Bicubic Sharper is generally considered a bad choice.
Cheers,
Bart
-
Many threads here on LuLa have discussed the much better algorithms that are available, including the ones in Lightroom. But Photoshop's Bicubic Sharper is generally considered a bad choice.
Yes, LR's is better for many reasons. Are you sure the OP is using that product?
Or perhaps the ultimate answer is to use a product from that company that promises to convert a JPEG to a raw, and allow one to edit the JPEG as if it were a raw?
Are there better interpolation algorithms? I'm sure there are, including LR. Does it make a big deal especially for images on a web page, where dog knows what kind or quality display is being used? I suspect not. I'd be far more concerned about all the images on the web being viewed without color management in a browser, or on displays without a lick of calibration than this nit picking about interpolations algorithms. The quality of images are all over the map. There are actual examples where Photoshop's Bicubic Sharper produces perfectly acceptable results. Even in the URL I provided. ;)
-
Yes, LR's is better for many reasons. Are you sure the OP is using that product?
Or perhaps the ultimate answer is to use a product from that company that promises to convert a JPEG to a raw, and allow one to edit the JPEG as if it were a raw?
Are there better interpolation algorithms? I'm sure there are, including LR. Does it make a big deal especially for images on a web page, where dog knows what kind or quality display is being used? I suspect not. I'd be far more concerned about all the images on the web being viewed without color management in a browser, or on displays without a lick of calibration than this nit picking about interpolations algorithms. The quality of images are all over the map. There are actual examples where Photoshop's Bicubic Sharper produces perfectly acceptable results. Even in the URL I provided. ;)
Have you looked at the OP's downsampled image?
Cheers,
Bart
-
Have you looked at the OP's downsampled image?
Cheers,
Bart
Have you looked at my comments about those images?
Yes I have!
-
Have you looked at the OP's downsampled image?
Cheers,
Bart
Have you looked at the OP's raw, or the high rez rendered from it to produce the downsampled image?
-
For the OP:
https://petapixel.com/2010/02/17/use-bicubic-sharper-for-web-resizing/ (https://petapixel.com/2010/02/17/use-bicubic-sharper-for-web-resizing/)
Thank you :)
-
Do i understand from this discussion that the latests versions of LR and Photoshop use different techniques for sharpening ?
and that LR does a better job?
-
Do i understand from this discussion that the latests versions of LR and Photoshop use different techniques for sharpening ?
and that LR does a better job?
Yes, always have.
-
Have you looked at my comments about those images?
Yes I have!
Could have fooled me, you even called it a 'tad soft'.
Besides that, you seemed content with the result and in defense of the software that produced that abominable result, which is strange, given the heavy artifacts. But, whatever makes you happy ...
Cheers,
Bart
-
Could have fooled me, you even called it a 'tad soft'.
Besides that, you seemed content with the result and in defense of the software that produced that abominable result, which is strange, given the heavy artifacts. But, whatever makes you happy ...
Cheers,
Bart
Easily fooled. :D
An assumption as well; I use LR. Very content!
As for further assumptions; you have seen the OP's raw and full max rendered image as asked ?
-
Easily fooled. :D
An assumption as well; I use LR. Very content!
Then why did you share a link to an article about Bicubic Sharper? How does that help the OP?
As for further assumptions; you have seen the OP's raw and full max rendered image as asked ?
I have not seen it, I did see a crop. Did you see the OP's Raw and full max rendered image?
Cheers,
Bart
-
Then why did you share a link to an article about Bicubic Sharper? How does that help the OP?
I have not seen it, I did see a crop. Did you see the OP's Raw and full max rendered image?
Cheers,
Bart
I linked it BECAUSE he use PS! Unless we hear otherwise it's fine but LR is SLIGHTLY visually better.
No, you haven't see the original data so don't make further assumptions about what you see posted here.
No I haven't seen it either and would never post assumptions about the resampled data FROM it!
-
I linked it BECAUSE he use PS! Unless we hear otherwise it's fine but LR is SLIGHTLY visually better.
No, you haven't see the original data so don't make further assumptions about what you see posted here.
No I haven't seen it either and would never post assumptions about the resampled data FROM it!
The result (unfortunately) looks like shit, no assumptions needed, just a single observation is needed.
Cheers,
Bart
-
The result (unfortunately) looks like shit, no assumptions needed, just a single observation is needed.
Cheers,
Bart
Your subjective opinion of what it looks like with massive assumptions as to WHY.
For all "we" know, the original is a camera JPEG. But from a raw, even worse; you have zero idea about the process of rendering it.
Unless you experience it, you are only imagining it. That results in credibility issues.
Enough said on that aspect of aiding the OP. ;)
-
Your subjective opinion of what it looks like with massive assumptions as to WHY.
For all "we" know, the original is a camera JPEG. But from a raw, even worse; you have zero idea about the process of rendering it.
Unless you experience it, you are only imagining it. That results in credibility issues.
Enough said on that aspect of aiding the OP. ;)
https://forum.luminous-landscape.com/index.php?topic=129534.msg1099940#msg1099940
The advice was clear ...
Cheers,
Bart
-
https://forum.luminous-landscape.com/index.php?topic=129534.msg1099940#msg1099940 (https://forum.luminous-landscape.com/index.php?topic=129534.msg1099940#msg1099940)
The advice was clear ...
Cheers,
Bart
More assumptions. You have absolutely no idea the condition of the original rendered image from the raw. Please, let the OP fully explain his workflow and let's advise him based on what he has and where he used it not just about the resampling that's a tiny part of an entire process.
-
I do not understand that LR and Photoshop, both from Adobe, use different sharpening techniques, while the Raw conversion is the same in both.
If it is clear that the methods used in LR are better, why aren't they implemented in Photoshop?
Is there some internal competiton ? Does it has to do with legal rights to the impkemented methods?
-
More assumptions. You have absolutely no idea the condition of the original rendered image from the raw. Please, let the OP fully explain his workflow and let's advise him based on what he has and where he used it not just about the resampling that's a tiny part of an entire process.
Well first of all I repeat that I agree with somebody here that the picture was not a great example. Again sorry about that
It is a picture 7360x4912 taken at ISO 3200 1/200 sec f/1.4 w/50mm NO STABILIZATION
The focus distance - to the point of the nose - was 0.71m.
That gives a DOF of 1.6 cm which could explain why the whiskers (and the rest) are soft (DOF 0.700 m - 0.716m)
My workflow:
Import the file to LR. In this case there were no adjustments, especially NO SHARPENING.
Export as TIF to PS
Resize to 800x534
Convert to 8 bits
Convert from ProPhoto to sRGB
Duplicate 2x (and save as three .jpg's: original - NIK - Photokit)
NO Capture Sharpening (wanted just a test on the behaviour when Output-Sharpening)
Sharpened with default values one w/ NIK Output Sharpener the other with PK (Superfine Edge Sharpen is the default)
That was it.
I opened all three, set the displays to 1:1 and arranged them vertically for examination.
I liked the unsharpened most, second came the PK and IMO Nik was the "most crunchy" of all.
I stress that I have no practical problems with that, since both versions let adjust themselves via layers
But I wanted to understand what happens and why and any possible drawbacks I am not aware of
Hope I didn't forget anything
Thanks for your patience :)
-
I do not understand that LR and Photoshop, both from Adobe, use different sharpening techniques, while the Raw conversion is the same in both.
If it is clear that the methods used in LR are better, why aren't they implemented in Photoshop?
Is there some internal competiton ? Does it has to do with legal rights to the impkemented methods?
The data processed (especially raw data), the encoding of the data, the color space and it's TRC are different in both along with the algorithms.
LR uses high bit linear processing for all data in a wide gamut color space, PS may not (usually not for linearity). LR uses an adaptive sharpening algorithm not found in PS. LR of course can process raw data, Photoshop cannot. The two are simply different in many respects.
-
The data processed (especially raw data), the encoding of the data, the color space and it's TRC are different in both along with the algorithms.
LR uses high bit linear processing for all data in a wide gamut color space, PS may not (usually not for linearity). LR uses an adaptive sharpening algorithm not found in PS. LR of course can process raw data, Photoshop cannot. The two are simply different in many respects.
Photoshop ACR processes raw data- i guess - that is the same engine as LR?
and the shapening procedures in it are the same? They read and like to process the same on the basis of the same XMP file.
If not i will not use photoshop ACR again...
-
Photoshop ACR processes raw data- i guess - that is the same engine as LR?
and the shapening procedures in it are the same?
If not i will not use photoshop ACR again..
Yes, LR uses the same ACR engine and one can move back and forth if the two are on version parity.
-
Question for Andrew:
In his post above kers is muddying the waters by asking about LR's Develop Module sharpening and ACR (as transferred in the xmp) although the thread has been about comparing LR's Export/Resizing and Output Sharpening vs. PS's Resampling and output sharpening from another plugin.
However, (although I suspect it is rarely used) ACR does have the ability to send to PS a resized and output sharpened image, as accessed from the Workflow Options dialog(center screen, bottom), whose interface strongly resembles LR's Export page.
Can we assume that Workflow Options and Export are the same in these functions (aside from the difference that W.O. does not write a file to disc)?
-
This subject matter definitely does not need sharpening. Modern cameras already produce images way sharper than in the film era.
Lightroom has output sharpening and enough choices (none, low, medium, high) to satisfy everyone. When I post to Flickr, I sometimes choose "none" or "low," as Flickr's algorithm tends to oversharpen by itself.
The extra sharpening of digital is one of the things I like and don't like about it. It's often too clinical looking and has the "soap opera" effect. On the other hand, sometimes the really sharp looks adds to the image. I know I have to add a lot of sharpening when I scan my medium format film shots. But things seem to blend better with film.
-
That setting is just fine. It will automatically select the Bicubic algorithm based on if you're sizing up or down. Nothing wrong with how PS does this! Users of course can hose any image by over sharpening when they shouldn't.
On my screen (a NEC PA271Q), the original looks a tad soft.
One major issue with sharpening anything visually is that as you move some sharpening controls up, feel it's a tad too much, then lower, the effect appears to look 'too soft' in comparison. It's one reason why people over sharpen. The eye is playing ticks on them. You need to walk away or refresh your view, then reexamine the image, of course at 1:1 (100%).
Interesting observation. Never knew that. Thanks for passing it on.
In my operation, I either use the cameras jpeg with no sharpening or let LR apply the initial "25" to RAW shots. Then, for both types, I use LR's "sharpen for the Screen" when I output the processed image file. I show my photos either on my 4K UHDTV, on the web, or on my desktop monitor.
When I scan film, I sharpen the heck out of it in LR so minimally does the scanner do its job. ( I scan flat).
-
I made three copies of this image and sharpened one with NIK Output and a second one with Photokit using their defaults for this size (800x540) and for the web
Both sharpening tools have (for my taste) oversharpened, one more than the other.
Is there a reason for that?
My experience with sharpening for Inkjet has been always satisfactory but not for the monitor display.
I know, one can adjust it but perhaps I am missing something?
2 and 3 look oversharpened to me.
-
Question for Andrew:
In his post above kers is muddying the waters by asking about LR's Develop Module sharpening and ACR (as transferred in the xmp) although the thread has been about comparing LR's Export/Resizing and Output Sharpening vs. PS's Resampling and output sharpening from another plugin.
However, (although I suspect it is rarely used) ACR does have the ability to send to PS a resized and output sharpened image, as accessed from the Workflow Options dialog(center screen, bottom), whose interface strongly resembles LR's Export page.
Can we assume that Workflow Options and Export are the same in these functions (aside from the difference that W.O. does not write a file to disc)?
Again, if on version parity and using the identical settings, ACR and LR produce identical rendering of the data.
-
..
In his post above kers is muddying the waters by asking about LR's Develop Module sharpening and ACR (as transferred in the xmp) although the thread has been about comparing LR's Export/Resizing and Output Sharpening vs. PS's Resampling and output sharpening from another plugin....
OK sorry for that....but became confused.
on topic: i use a combination of making the image smaller in a few steps -bicubic standard and adding some low sharpening here and then; it works well for me. I made it into an photoshop action.
so for instance step1 : little sharpening- 8000px to 3940px - step2 little sharpening 3940px to 1920px.. export etc
-
In case somebody could find this useful:
I sent Bart a file (again my cat) in original size.
He returned it to me downsized to 800x together with a sharpened copy for comparison
1. original2_800x.jpg is downsampled from the original file with a method that's very similar to the one used in Lightroom.
https://www.dropbox.com/s/xkjsok4a8gjncxd/original2_800x.jpg?dl=0
2. original2_800x-sharpen-030-010-010.jpg is that same file, with some sharpening added (I used Topaz Sharpen AI version 1.1.0, in Sharpen mode).
https://www.dropbox.com/s/67kco5s7k9oravc/original2_800x-sharpen-030-010-010.jpg?dl=0
I sharpened the original with the default values (RemoveBlur .50 Supress Noise .50 Add Grain 0) of Topaz Sharpen AI version 1.1.0, in Sharpen mode
The result was (for my taste) oversharpened.
I repeated this using the values from Bart (RemoveBlur .30 Supress Noise .30 Add Grain 10) And obtained the same result (of course) as Bart, a pleasant sharpening.
Using Photokit w default values for Output Sharpening produces a similar (a tad sharper) result
-
Rabanito's original full size image is an image that is very sharp, but with shallow Depth of Field, and with some noise.
As was obvious to me from the start, the original reduced size image did suffer from the process of downsampling, with ringing and aliasing/stairstepping artifacts as the tell-tale signs. So some improvement in that initial step was possible. Using a better downsampling method, produced a more robust image for further processing/sharpening.
The amount of subsequent sharpening remains a personal preference thing. But the sharpening method should IMHO preferably be either deconvolution sharpening (Focusmagic is still a great tool for that), to restore some of the resampling blur or, as it has started becoming available, detail replacement by AI.
Deconvolution, or AI, both allow minimizing the risk of amplifying the artifacts that were already there, or of introducing new artifacts.
It was an interesting exercise.
Cheers,
Bart
-
I think what's happening is that trying to correct a too narrow DOF with sharpening just doesn't work. You have to get it right in the camera first.
-
I think what's happening is that trying to correct a too narrow DOF with sharpening just doesn't work. You have to get it right in the camera first.
The intention was not to "correct" the DOF but "to restore some of the resampling blur", quoting Bart.
The DOF is OK w me ;)
-
Look at the cat's body in the different pictures and you will see some where all the detail is missing. I would start by doing a Levels type of adjustment and then adjusting the contrast of the full image and then do a resizing for the Web and lastly to sharpen the image. The smaller the Web file size the more sharpening I need to use.
That is why I always start with the original file, resize for the output file size needed, and then sharpen. I do not resize to 1000 x 1000 and sharpen and then resize that file to 400 x 400 and sharpen again as the resulting image quality will be much less.