Super Resolution

[BobT]

Lightroom Version Number

10.3

Operating System

1. Windows 10

I think this was covered before but now that we have Super Resolution in LrC proper, a recap might be good. I was wondering when is the best time in PP to do it? It seems to me logically after cropping but before sharpening might be best. I know, I should experiment myself but it would be a waste not to take advantage of the blood and sweat of the gurus here.

 

[Johan Elzenga]

Remember that Lightroom adjustments are non-destructive. They do not alter the original image in any way, and so they do not influence the outcome of Super Resolution.

 

[BobT]

OK, I see. I was looking at it more like another though powerful Develop Module feature. It's not that at all. I guess you've answered my question. Thanks.

 

[chrisbellamy]

It would be nice if the 'enhanced' DNG could be optionally generated just for the crop, to keep file size down where the target is small and we are trying to make the most of a suboptimal number of pixels on target (as in some of Eric Chan's examples); the original RAW would be intact of course.

 

[clee01l]

It would be nice if the 'enhanced' DNG could be optionally generated just for the crop, to keep file size down where the target is small and we are trying to make the most of a suboptimal number of pixels on target (as in some of Eric Chan's examples); the original RAW would be intact of course.

This makes sense. I have some very large 48 mp (~60MB) NEFs the result is a ~640MB DNG.

 

[Linwood Ferguson]

Remember that Lightroom adjustments are non-destructive. They do not alter the original image in any way, and so they do not influence the outcome of Super Resolution.

That's true, but they are re-applied (and thus recalculated) to the after-enhanced image. For example, if I take a brush stroke that is 100pixels wide, after I enhance it lightroom will turn that into one 200 pixels wide to "match". Now if I were painting something in, adjusting fine detail, I might actually do more strokes more narrow now that I have a bigger canvas.

Things like sharpening I suspect are similar to the extent a human is making a decision. So if you sharpening zoomed 1:1 until it looks crunchy (whatever your definition of that is), would you stop earlier or later if twice the resolution was zoomed 1:1 and you were moving the slider?

I certainly agree with Johan that they don't change the outcome of the super resolution itself. It doesn't matter. The super-resolution is done without regard to them. BUT... where a human is involved, that human may do them differently if dealing with twice the pixels; just doubling the linear size of local adjustments, and applying the same values to other slides, may not quite look the same with twice the resolution.

I think I would recommend that things you apply by eye (as opposed to a preset, or just "I ways use 3 for that") are best done after super resolution, so your eye benefits (or at least adapts to) the added resolution. Especially local adjustments, sharpening, Clarity, Texture, Dehaze, maybe Aberration corrections. Others, like Color, white balance, any grading changes I would think end up the same.

 

[prbimages]

It would be nice if the 'enhanced' DNG could be optionally generated just for the crop, to keep file size down where the target is small and we are trying to make the most of a suboptimal number of pixels on target (as in some of Eric Chan's examples); the original RAW would be intact of course.

You may already know this, but you can apply Super Resolution to other file formats such as TIF, PSD, JPG. So you could save your cropped RAW image as (say) a TIF file and then run Super Resolution on that smaller file.

 

[Johan Elzenga]

You may already know this, but you can apply Super Resolution to other file formats such as TIF, PSD, JPG. So you could save your cropped RAW image as (say) a TIF file and then run Super Resolution on that smaller file.

You can, but the result may not be as good, unless you take a two step approach. If you apply super resolution on a raw file, then ‘enhance details’ is applied first. If you apply super resolution on a tiff file, you would have to do this first yourself before you create a tiff from the cropped raw file.

 

[clee01l]

You may already know this, but you can apply Super Resolution to other file formats such as TIF, PSD, JPG. So you could save your cropped RAW image as (say) a TIF file and then run Super Resolution on that smaller file.

I think since Adobe is creating a derivative file in the RGB DNG, it should honor any LR adjustments after creating the Super Resolution. And this should include any crop.

From what I see in my one Super Resolution effort, is that it is no different from Topaz GigaPixel. Also I could discern no improvement between the SuperResolution image and a Zoomed version of a Lightroom Edited RAW.


Sent from my iPad using Tapatalk

 

[BobT]

I think since Adobe is creating a derivative file in the RGB DNG, it should honor any LR adjustments after creating the Super Resolution. And this should include any crop.

From what I see in my one Super Resolution effort, is that it is no different from Topaz GigaPixel. Also I could discern no improvement between the SuperResolution image and a Zoomed version of a Lightroom Edited RAW.


Sent from my iPad using Tapatalk

Can you please elaborate on your last sentence? I'm probably not reading it correctly but it implies to me that you are not too impressed with Super Resolution.
Regarding Topaz GigaPixel, last time I looked, for best results, it relied on online access to a vast Topaz AI database. I don't see anything like that in Super Resolution.

 

[clee01l]

Can you please elaborate on your last sentence? I'm probably not reading it correctly but it implies to me that you are not too impressed with Super Resolution.
Regarding Topaz GigaPixel, last time I looked, for best results, it relied on online access to a vast Topaz AI database. I don't see anything like that in Super Resolution.

As I prefaced my comment, I have only tested this with one image. And you are correct I've not been too impressed. After I have tried this on several images, I might change my mind. However, until I have the item to fully test this, I remain unimpressed.

 

[BobT]

As I prefaced my comment, I have only tested this with one image. And you are correct I've not been too impressed. After I have tried this on several images, I might change my mind. However, until I have the item to fully test this, I remain unimpressed.

I tend to agree. For nature photography, I see not much improvement over upsampling. Built environments have many straight lines and some smooth curves. AI might readily detect this and would be safe to assume that lines and curves should be sharp. Not so for nature. No straight lines and few smooth curves and none can be assumes to be sharp. A more sophisticated AI might be able to detect animal shapes then sharpen hair and feather detail accordingly but not so LR just yet.

 

[tspear]

I was curious so I played with it on some older images. I was afraid to see the size of the results on my new Sony with 61mb files .
It actually was a mixed bag. On people and man made stuff it did well. With plants, cats and related it did very poorly.
Here are the general type of images I used for testing:
1. Family photo
2. Family at holiday diner table
3. Family in front of tree in front of house
4. Cat on deck
5. Cat at base of tree
6. Almost macro shot of flower
7. Almost macro shot of leaf
8. Long exposure shot of beach

When the photo had nature plus people/man made it made the image really weird. The difference in detail was jarring. If just people and almost no nature it worked well. I might consider it for some really old images or maybe for some phone images, otherwise I would not likely use it.

Note: I have no idea how long it takes. I would start the process on an image in the morning or evening and let it run.

Tim

Sent from my HD1907 using Tapatalk

 

[BobT]

I was curious so I played with it on some older images. I was afraid to see the size of the results on my new Sony with 61mb files .
It actually was a mixed bag. On people and man made stuff it did well. With plants, cats and related it did very poorly.
Here are the general type of images I used for testing:
1. Family photo
2. Family at holiday diner table
3. Family in front of tree in front of house
4. Cat on deck
5. Cat at base of tree
6. Almost macro shot of flower
7. Almost macro shot of leaf
8. Long exposure shot of beach

When the photo had nature plus people/man made it made the image really weird. The difference in detail was jarring. If just people and almost no nature it worked well. I might consider it for some really old images or maybe for some phone images, otherwise I would not likely use it.

Note: I have no idea how long it takes. I would start the process on an image in the morning or evening and let it run.

Tim

Sent from my HD1907 using Tapatalk

With my old clunker of a PC, it takes less than half an hour but I have much smaller m4/3 files. However, I dare not touch the PC while it's running Super Resolution or it crashes. My system has close to minimum recourses so not unexpected.