D800 and the effort of down sampling

Triggaaar

Suspended / Banned
Messages
1,456
Name
Mike
Edit My Images
Yes
The new D800 has been seen primarily as a studio and landscape model, but there are signs that it could be quite the all-rounder, capable of decent results at high ISO too. However, whenever high ISO comparisons are made, should we be comparing 100% crops, or should we be comparing prints, and is extra work required for the D800: are we comparing apples to apples?

Whenever I export a picture from Lightroom, I either specify print dimensions and select 300 dpi, specify the number of pixels along the long edge for computer displaying, or select a quality of jpeg (high, low, medium etc).

So in what I assume is a typical workflow, what would we do differently if moving from a lower resolution model to a D800? When people are saying (in the context of high iso pictures) ‘the D800 image only looks as good/better if you down-sample’, what is this extra step they’re having to perform? Why can't you take a picture in raw, edit as usual, and export a jpeg for printing, and that be it?
 
Why can't you take a picture in raw, edit as usual, and export a jpeg for printing, and that be it?

You can, it is only the pixel peeping fanboy/detractors that say otherwise.

just enjoy your camera and shoot!
 
You can, it is only the pixel peeping fanboy/detractors that say otherwise.
Well either I'm missing something fundamental (like a brain, which isn't out of the question) or people (decent photographers) are just making stuff up.

just enjoy your camera and shoot!
I know that's generally right, but if we're charging for our shots we need to be critical of our work (ie, the finished output). But I'm not reading this as 'A is as good as B unless you're fussy', I'm reading it as 'the D800 isn't as good at high iso unless you go to all this extra effort of down-sampling to make it so'. What is this extra work?
 
I have seen samples from the D800 and the 5d3 and they both work very well, what more do you need. Sometimes it is user error that causes these problems.
 
....Why can't you take a picture in raw, edit as usual, and export a jpeg for printing, and that be it?

That's how I work for the magazines at work, always have. The printing process will see to it that IQ only reaches a certain ceiling. If we had the benefit of better paper stock and a printer who was geared towards producing a much higher spec product, then I would certainly look at how my workflow would need to change to harness the benefits.:)
 
this is what i was saying on the d800 thread and got jumped on for.

if you process as normal the results will NOT be as good as a d4 or nearly as good as everyone is making out.
even the dxo review/test states this.
they reduce all sensors to 8mp for reviewing.
http://www.dxomark.com/index.php/Pu...d-computation-of-DxOMark-Sensor-normalization

But if u downsample to 12mp then there is a vast inprovement.
heres a link to down sampling http://mansurovs.com/how-to-properly-resize-images-in-photoshop
 
The new D800 has been seen primarily as a studio and landscape model, but there are signs that it could be quite the all-rounder, capable of decent results at high ISO too. However, whenever high ISO comparisons are made, should we be comparing 100% crops, or should we be comparing prints, and is extra work required for the D800: are we comparing apples to apples?

Whenever I export a picture from Lightroom, I either specify print dimensions and select 300 dpi, specify the number of pixels along the long edge for computer displaying, or select a quality of jpeg (high, low, medium etc).

So in what I assume is a typical workflow, what would we do differently if moving from a lower resolution model to a D800? When people are saying (in the context of high iso pictures) ‘the D800 image only looks as good/better if you down-sample’, what is this extra step they’re having to perform? Why can't you take a picture in raw, edit as usual, and export a jpeg for printing, and that be it?

If you're printing at less than 24x16 at 300dpi you are already downsampling.

Images you see on the web are going to be downsampled anyway unless they're less than 2MP.

The only reason people have said that you should downsample is that it stops people doing idiot 100% crop comparisons for different resolution sensors.

You can do exactly what you do now, just resizing for print size and you'll get the full benefit. The only thing that might change is how you downsample (the different algorithms have different noise/detail impacts).
 
this is what i was saying on the d800 thread and got jumped on for.

if you process as normal the results will NOT be as good as a d4 or nearly as good as everyone is making out.
even the dxo review/test states this.
they reduce all sensors to 8mp for reviewing.
http://www.dxomark.com/index.php/Pu...d-computation-of-DxOMark-Sensor-normalization

But if u downsample to 12mp then there is a vast inprovement.
heres a link to down sampling http://mansurovs.com/how-to-properly-resize-images-in-photoshop
Thanks for that second link, interesting read. But it does confuse the issue. He's going to extra lengths to down-sample, that I don't think we need.

I think you're misunderstanding the process. You say
"if you process as normal the results will NOT be as good as a d4 or nearly as good as everyone is making out."
Who here puts a picture from their D700 (or whatever) without down-sampling? As I've said above, I simply Export my pictures through Lightroom, and that automatically does the down-sampling, regardless of whether I'm using my D700 or a D800.

No one ever sees my pictures unless I've exported them from Lightroom. Why would they? So all this talk of 'you need to down-sample your D800 high iso images to make them as good as a D700' is nonsense, right?

If you're printing at less than 24x16 at 300dpi you are already downsampling.

Images you see on the web are going to be downsampled anyway unless they're less than 2MP.
That's what I thought, thanks.

The only reason people have said that you should downsample is that it stops people doing idiot 100% crop comparisons for different resolution sensors.
Unfortunately you're probably talking about the minority. Most comparisons I've seen involve people say 'why buy a 36mp camera to down-sample, it's a waste of time, I'd rather have my 12mp D700'. They appear totally oblivious to the fact their D700 is down-sampling too.

The only thing that might change is how you downsample (the different algorithms have different noise/detail impacts).
You mean that for critical work (say a large print for a client) I might want to go to PS instead of exporting from Lr? I'd hoped that Lr would be just as good.
 
Last edited:
thanks not some thing i new i was already doing or should be doing.
im sure my pics are out putted at orgional size :O

if u click my flickr pics and go to large size there huge .you have to scroll across the screen. o_O

learn sonething new everyday
 
Last edited:
People who keep on about doing pixel level comparisons just confuse me. I don't know about them, but I print and view images not pixels. I like showing off my best photos, but rarely if ever have I shown someone a pixel and said "this is my favourite pixel, a classic example of RGB(201,56,123), the classiest of pixels I'm sure you'll agree."

Compare the output of the cameras at the size(s) you'll most often use. Anything else is pointless.
 
You mean that for critical work (say a large print for a client) I might want to go to PS instead of exporting from Lr? I'd hoped that Lr would be just as good.

The different resampling methods retain different frequencies of detail and noise. So sometimes a naive downsampling will not get quite the results expected. A lot of those cases can be resolved by simply running a weak gaussian blur filter prior to downsampling.

Any sharpening after that can be done to taste. The downsampling algorithms are identical between Lightroom and Photoshop. There's no need to use PS rather than Lightroom for sharpening afaik.


The only counter argument would be if you did actually have to do this:

(link to down sampling as posted by pmac) http://mansurovs.com/how-to-properly-resize-images-in-photoshop


Is it correct to assume we don't need to do that, the down-sampling from Lr is good?

That link is odd, he writes things about downsampling that have nothing to do with downsampling.

Here are the actual steps he does when downsampling (the rest is normal image processing he would do anyway):

Down-sample the image to target resolution (target resolution is typically 1024 pixels wide for horizontals and 500 pixels wide for verticals)

Run another pass of sharpening at 25-50% (Unsharp Mask)

Export the image in sRGB color profile

He could do with a good copy editor :)
 
Last edited:
The different resampling methods retain different frequencies of detail and noise. So sometimes a naive downsampling will not get quite the results expected.
Understood, although I'd have guest that it's software on sites like Flickr that give poor down-sampling, not applications like Lightroom?
 
Last edited:
Understood, although I'd have gets that it's software on sites like Flickr that give poor down-sampling, not applications like Lightroom?

It's not so much poor as that they suppress and allow through different frequency components (I'm only talking about LR and PS type downsampling, as they're the ones you can control - Bicubic sharper/smoother etc).

Essentially it's almost a complete non issue. The rare occasions it is an issue for noisy images you can fix it by either using one of the other algorithms or running a weak gaussian blur over the image before downsampling. Or run a pass of noise reduction (but that's essentially blurring anyway).
 
The only counter argument would be if you did actually have to do this:

(link to down sampling as posted by pmac) http://mansurovs.com/how-to-properly-resize-images-in-photoshop


Is it correct to assume we don't need to do that, the down-sampling from Lr is good?

Unless your current camera precisely matches the required pixel dimensions of your current (print size x dpi / web size / whatever) with a 1-to-1 mapping, I don't see how that is any different to whatever camera you are currently using. Most people resample from camera to output, whether they realise it and do it explicitly or not.
 
Essentially it's almost a complete non issue. The rare occasions it is an issue for noisy images you can fix it by either using one of the other algorithms or running a weak gaussian blur over the image before downsampling. Or run a pass of noise reduction (but that's essentially blurring anyway).
Thanks for trying to explain, although I'm not totally understanding.

If I want to use a picture with high iso (lets say 6400 as an example), I will always apply some noise reduction in Lightroom. With a little cropping and a D700 that may result in very small up or down-sampling (if I'm exporting to print). If I were to use a D800 in the same situation, and allow Lightroom to down-sample for me when exporting the output, would the end picture be better/worse (I'm not expecting you to know, but that's essentially the question we're all interested in). Or would we have to change our workflow - would we have to go to PS to do some more clever down-sampling? (in fact, should we be doing that already with our 12mp files?)
 
Thanks for trying to explain, although I'm not totally understanding.

If I want to use a picture with high iso (lets say 6400 as an example), I will always apply some noise reduction in Lightroom. With a little cropping and a D700 that may result in very small up or down-sampling (if I'm exporting to print). If I were to use a D800 in the same situation, and allow Lightroom to down-sample for me when exporting the output, would the end picture be better/worse (I'm not expecting you to know, but that's essentially the question we're all interested in). Or would we have to change our workflow - would we have to go to PS to do some more clever down-sampling? (in fact, should we be doing that already with our 12mp files?)

It should be better. :)
 
Back
Top