Post processed image differs in Library & Develop Modules

[MMarz]

Working on a few images from a preserve shoot and after processing one in LR2, I notice that the sharpening appears far greater when the image is viewed in the Library module versus how it looks in the Develop module...

In the Develop Module..

Same image when viewed in the Library module

 

[Sean McCormack]

Develop only shows Sharpening at 1:1. Going back to Library and rendering the 1:1 there will then show that actual shot.
You could of course make a feature request, but as Develop renders on the fly, you could end up doing an awful lot of waiting on rendering in Develop.

 

[MMarz]

Wow.. that makes no sense to me at all.

I understand the need for us to go to 1:1 to apply sharpening, but why wouldn't LR then show the sharpening applied to all views within the Develop module so you can see exactly the same image that you will see elsewhere withing LR, as well as when you output??

Holy smokes...this totally changes my work flow..and maybe my opinion of LR.

So, if I work through an image, and end with sharpening, I I see all the steps I apply to crop, exposure, tone, localized edits..but not sharpening? What on earth is the reasoning for this??

 

[Brad Snyder]

Michael, the best I can tell you is to paraphrase the guy that wrote the code.

Develop rendering is demand-driven, if you aren't at 1:1, then you don't get 'detail' processing. This is a performance issue.

Library assumes static images, Develop assumes dynamic changing images. Library 1:1 previews are sharpened and NR'd, because the the image is presumed to be static, corresponding smaller previews are down-sampled from that, so they show the sharpening and noise-reduction.

So, if I work through an image, and end with sharpening, I I see all the steps I apply to crop, exposure, tone, localized edits..but not sharpening?

You can see it in 1:1/greater in Develop, and get a static rendering of it in Library at all zooms.

What on earth is the reasoning for this??

Performance optimization

 

[Braders]

Wow.. that makes no sense to me at all.

I understand the need for us to go to 1:1 to apply sharpening, but why wouldn't LR then show the sharpening applied to all views within the Develop module so you can see exactly the same image that you will see elsewhere withing LR, as well as when you output??

Some would argue that you can really only see the sharpening at 1:1, and that standard preview size you will not see a significant difference.

 

[MMarz]

Some would argue that you can really only see the sharpening at 1:1, and that standard preview size you will not see a significant difference.

I'd argue that you absolutely can..my down-res'd images posted above are proof... They are at Fit size...Can you see a difference?

 

[MMarz]

Michael, the best I can tell you is to paraphrase the guy that wrote the code.

Develop rendering is demand-driven, if you aren't at 1:1, then you don't get 'detail' processing. This is a performance issue.

Library assumes static images, Develop assumes dynamic changing images. Library 1:1 previews are sharpened and NR'd, because the the image is presumed to be static, corresponding smaller previews are down-sampled from that, so they show the sharpening and noise-reduction.

I can understand the performance logic... but I can't understand how we are to process an image, if the preview we see is different in each module. I guess I am just unable to rationalize this. Obviously if you can't back out of a 1:1 view in the Develop module and see what your edits have done at what might be an output size, how do you know what th end result is?? It almost a guarantee that you will either over-sharpen or have to switch back and forth between modules...now that is a performance issue.:evil:

 

[Braders]

I agree the changes can be seen, but i have never seen such an appreciable difference at fit view than the example you give.:?

 

[MMarz]

I agree the changes can be seen, but i have never seen such an appreciable difference at fit view than the example you give.:?

That was with Matt's Landscape Sharpening, minus 14...

Amount 83
Radius 1
Detail 9'
Masking 15

 

[ernie]

I would agree with Michael that it seems rather silly to have different views in different modules. But that aside-what about develop presets (which include added sharpening) that are applied at import. Does that sharpening show up since it was applied already to the image as opposed to sharpening applied during the develop process? I'm thinking it does show on mine, but I have been known to be wrong a time or two, sooo...?

 

[Brad Snyder]

But that aside-what about develop presets (which include added sharpening) that are applied at import. Does that sharpening show up since it was applied already to the image as opposed to sharpening applied during the develop process? I'm thinking it does show on mine...

Ernie, I don't think so. Only at 1:1 or larger.

There's really nothing magic about a preset, it's basically just a list of slider settings.


On a tangent, for the educational value of it, you might like to examine an XMP sidecar file with a text editor. You can readily see, it's a list of instructions of how to set all of the develop mode sliders, and a list of the applied metadata, etc. expressed in predetermined format. A develop preset looks very similar when examined.

 

[ernie]

OK Brad, thanks. I just wondered since the default settings, I assume, show up in the newly opened file if any presets were applied at the same time as the default settings. Sort of like making them the "new" default.
Lightroom has it's quirks like a lot of other software I guess. But the good FAAAR outweighs the bad.
I'm really not that good at running computers. I understand about being just a set of instructions, but I don't even know how to begin to use a text editor to do anything. I'm not even sure what a text editor is. I think notepad is one, but I've never used it. I'm like a lot of folks I think. I know the things I do, but if I get beyond that the fog rolls in.

 

[Braders]

You know, i just went through a folder of mine shot in the Californian Redwoods and i am quite shocked.

What i am seeing is a whole series of images with a lot of fine detail, green foilage, twigs, bark etc, that i know i have applied sharpening at 1:1 in develop previously, that on first glance appear unsharpened.

I then zoom in to 1:1, and zoom out to standard preview and the difference is stark.

mmmm, as i remember, this was not the case in 1.4.1 for the develop or library modules, BUT was the case for the slideshow. Previously in the slideshow, one had to ensure 1;1 previews were generated in order to see the sharpening and noise reduction (at standard view) in the rendered slideshow.

I hope i don't have to have all my images rendered at 1:1 to ensure optimal standard size viewing?

 

[MMarz]

I hope i don't have to have all my images rendered at 1:1 to ensure optimal standard size viewing?

If that's what it would take to allow all views to be consistent I would do it.

I'm still so shocked with this I haven't been able to revisit it with a clear head.. so that I can verbalize and then understand how to proceed. I'm just can't phathom how it this is supposed to fit in you r workflow.

You adjust an image's tone curve, exposure, clarity etc in the Develop module and it's WYSIWIG... Once you're done and you want to apply final sharpening you do it at 1:1 but when you go back to a standard preview, still in Develop, so you can admire your work. you don't get the full deal...you can't see how much sharpening was applied. It's so easy to ruin a digital image with too much sharpening. Still you have to exit to the Library or output the image to see how much good or damage you've done.

How is it that only a few people feel this isn't OK? I'm left thinking there must be some flaw in my work-flow. I must not understand this well enough...

 

[Brad Snyder]

Well, let's think about this for a minute.

What is sharpening? Fundamentally it's an adjustment of the contrast of edge pixels. In many cases we're talking about about a region of 2 or 3 pixels in width, isn't that what the radius setting is about? Seriously, zoom in to 8:1 or more and see the sharpening pixel by pixel. It's a tiny little tweak, with large perceptual result.

Now, take your 8, 1', or 12+ megapixel image, and display it in fit/fill on your 1, 2 or 3 megapixel monitor, i.e. for every pixel in your image, throw 3 - 8 other pixels away.

I think it's amazing that the downsampling routines show any sharpening at all in fit/fill. I guess that's a testament to the high-frequency, edge-detection. I don't really expect it to accurately display sharpening at any resolution below 1:1.

That's what 1:1 means: show me all the pixels. In this instance, the zoom ratios of the display aren't so much about the viewable area of the image, but the mapping of image pixels to display pixels. Sure you can ask to see all of the image in fill/fit view, but it's just a polite fiction, until you can put your hands on a 12 megapixel monitor.

And realizing this, the design team elected not to throw away valuable processor cycles on real-time updating of an inaccurate representation.

Then add output/print sharpening, which is a black box whether done in LR or by the print lab.

A reduced scale display of an image on your monitor is really only a accurate indicator of how a reduced scale display of an image will appear on your monitor. For any other output, it's just educated guesswork.

But, quoting comic Dennis Miller, "That's just my opinion, I could be wrong." And if so, I welcome the opportunity to become educated.

 

[MMarz]

First, I would dare to be so arrogant as to say YOU are wrong. In fact, I understand 1''% of what you have said thus far, and since I have been on a quest for faster performance out of LR (and now have it) I would be out of line to challenge anything that would reduce performance.

However, here's my issue...

I don't really expect it to accurately display sharpening at any resolution below 1:1.

The fact is, sharpening IS displayed in a resolution below 1:1..in the Library mode. My struggle is limited to the inconsistency across views. Potential for a performance hit aside, I really don't understand how everyone applies any level of sharpening in the develop mod. It's like doing it with your eyes closed and my two images are a clear depiction of this. Every other adjustment made in the dev mode is displayed in the Fit view so you can see the results...but you can't effectively preview sharpening.

Maybe I need to sit with someone to watch their work flow.

I've watched every Killer Tip, YouTube and other video I could find on LR sharpening and though it always said that you must be in 1:1 to apply sharpening, nothing has ever mentioned the part about crossing your fingers when you THINK you are done with an image...

I get the feeling I am not making myself clear, simply because it is so obscure of an concept to me to not be able to really see the effect of an adjustment that I can't verbalize it.

I have to find a local LR workshop, or pay someone to help me work through this...

 

[Brad Snyder]

Michael, I'm with you, if anybody can explain it better, I really want to know the right answer. This subject just tends to sink into the black hole of debate. At this point, I'm just thinking out loud, who knows if I'm correct? I like it much better when the questions are "What button do I push to ....?"

The fact is, sharpening IS displayed in a resolution below 1:1 in Library

And that's where we disagree. Sure, 'a' representation is shown, but is it accurate? Again, who knows? But, physics dictates that most of the pixels are missing from down-sampling. Sure, your two images are different, but does either truly reflect the underlying pixels? I think they're nothing but a fairly rough simulation of what to expect from the final real-world output device(s). And unfortunately, in terms of both time and money, that's the only true test; print it out, or post it, and see.

In point of fact, I whole-heartedly agree with what you're asking for, I just don't think it's attainable with current display technology.

I've read basically two types of articles about sharpening. There are the ones about the pixel pushing, and the matrix math, and the yada... yada.... And then there are the ones by the 'artistes'. The latter all seem to come to this point: "There is no one correct setting, there's no magic formula, you just have to do it, observe the results, and readjust, and try, try again, 'til you get it right. In the end, you have to rely on past experience."

I get the feeling I am not making myself clear ....

No, I think we're reading you loud and clear. As I said above, and I'm not going to rewrite this again to put it back in logical order, I agree with you. I think you should be able to, I just don't see how you can.

(And I'm hopeful that you take all of this in the spirit in which I intend it, as a dialog between colleagues trying to get to the bottom of a vexing problem.)

 

[MMarz]

I just re-read my prior post...didn't read yours yet.. In any event..mine is supposed to say...."I would NOT dare to be so arrogant as to say YOU are wrong."

Now I'll take me head out of my butt and read your response. sorry about the fat fingers I have.

 

[Braders]

Michael, I'm with you, if anybody can explain it better, I really want to know the right answer. This subject just tends to sink into the black hole of debate. At this point, I'm just thinking out loud, who knows if I'm correct? I like it much better when the questions are "What button do I push to ....?"



And that's where we disagree. Sure, 'a' representation is shown, but is it accurate? Again, who knows? But, physics dictates that most of the pixels are missing from down-sampling. Sure, your two images are different, but does either truly reflect the underlying pixels? I think they're nothing but a fairly rough simulation of what to expect from the final real-world output device(s). And unfortunately, in terms of both time and money, that's the only true test; print it out, or post it, and see.

In point of fact, I whole-heartedly agree with what you're asking for, I just don't think it's attainable with current display technology.

I've read basically two types of articles about sharpening. There are the ones about the pixel pushing, and the matrix math, and the yada... yada.... And then there are the ones by the 'artistes'. The latter all seem to come to this point: "There is no one correct setting, there's no magic formula, you just have to do it, observe the results, and readjust, and try, try again, 'til you get it right. In the end, you have to rely on past experience."

No, I think we're reading you loud and clear. As I said above, and I'm not going to rewrite this again to put it back in logical order, I agree with you. I think you should be able to, I just don't see how you can.

(And I'm hopeful that you take all of this in the spirit in which I intend it, as a dialog between colleagues trying to get to the bottom of a vexing problem.)

Brad

It is intriguing for me. For the last year i have seen exactly what the experts describe, that in order to see any meaningful sharpening, the image needs to be viewed at 1:1. I tried on numerous occasions to see what effect the applied capture sharpening had on the standard size preview, and struggled to see any discernable difference.

When it came to noise on high ISO files, that was different. IF any high ISO file had not been rendered at 1:1, the noise was visibly present in both the library and slideshow. Once 1:1 was generated AND 'save to xmp' had been done, the slideshow images had there noise reduction rendered.

But, last night is the first time i have seen a noticable difference in sharpening between an image 'before' and 'after' 1:1 generation.

NOW, i have just got a new 26 inch monitor so that really made things obvious. However, on my 17"laptop, i still can see this difference, where as in 1.4.1 i could not.

Some of what you stated above i don't have a full appreciation for, but thats why i am in here.

Still learning computers!

 

[Mark Sirota]

I'm not going to defend Lightroom's methods, but what's happening is pretty simple to explain. There are two key points:

(1) When Lightroom generates previews, it picks a large size preview to generate, and then downsamples that to various other smaller sizes.

(2) Lightroom only applies noise reduction and capture sharpening if the large size preview is a 1:1 preview. This is because noise reduction and sharpening are time consuming, and downsampling dramatically reduces their importance. Perhaps someday when the typical computer is considerably more powerful than today's, this decision will be changed.

So if you want noise reduction and sharpening applied to your standard-size preview, you need to generate a 1:1 preview. It will then be downsampled to a standard-size preview. You can throw away the 1:1 preview after this if you prefer.

All that said, nothing about any of this changed between 1.x and 2.'.

 

[MMarz]

I think I recognize where my struggles are. I don't know enough about what happens back-stage in an application. Frankly, I don't really want to. There are certain aspects of Post that I enjoy, but most I do not. I am all about being out in the field capturing memories.

Sure, your two images are different, but does either truly reflect the underlying pixels?

You have hit the nail on the head here....Here-in is the extent of my struggle. I have two different renderings of my post work. That just doesn't make sense. I'm less concerned with how or why it happens and far more concerned THAT it is happening and that I'm not sure how to incorporate a remedy into my work flow.

"There is no one correct setting, there's no magic formula, you just have to do it, observe the results, and readjust, and try, try again, 'til you get it right. In the end, you have to rely on past experience."

By far, a better fit for describing my orientation!!

(And I'm hopeful that you take all of this in the spirit in which I intend it, as a dialog between colleagues trying to get to the bottom of a vexing problem.)

Don't think twice!! I am totally on board! If I could just stand over your shoulder while you run a shoot through post, many of my questions would probably be answered. hhhmm...wonder if the Geek Squad knows anything about LR?? :shock::shock: (not calling you a geek or anything like that!! :cheesy::cheesy

 

[Braders]

All that said, nothing about any of this changed between 1.x and 2.'.

I trust your statement here Mark, and by no means want to argue it, but i am seeing things differently.

Regarding sharpening only - I had not seen any appreciable difference until 2.'. Now it is quite suprising. Maybe, i wasn't paying enough attention.

 

[MMarz]

So if you want noise reduction and sharpening applied to your standard-size preview, you need to generate a 1:1 preview. It will then be downsampled to a standard-size preview. You can throw away the 1:1 preview after this if you prefer.

OK then.

This is the workaround, and I thank you!!

I will happily offer up the extra memory, and a bit of additional time at import, and render the 1:1's.

hhhmmm... wonder if there is better way. Maybe I should import with Standard previews, do an initial cull of the obvious rejects, apply some rating info and the generate the 1:1's on only those images that rate highly.

Thank you for stepping into this thread!!

 

[Ben Jenssen]

Ive been wondering about this since I started using Lightroom a year ago.

My biggest complaint is that when I do noise reduction in 1:1 view, when I zoom out, the preview doesnt change.

When doing quite heavy NR, this is very frustrating. The noise comes back when I zoom out! And I cant find a way to force a re-render of the preview, with NR, in develop mode, while working. I should make a feature request...

Edit: And with my new blazingly fast Mac Pro, the processor speed issues does not concern me.
LR 2.1 is just so fast and responsive, gradients and everything. A quantum leap from my G4 single 1.8gHz.

Another edit: I Just found out: I can go from develop to loupe, zoom to 1:1, zoom out, the preview reflects the NR in full frame, but when I return to develop, the noise comes back. But I would call it a workaround.