Super Resolution

[Zenon]

Operating System

1. macOS 15 Sequoia

LrC 14.5.1

I've never a need for it but I thought I'd try it out. I have a file that is 6000 x 4000 and I export it at as a Jpeg @1600 on the long side. I reimported it as a TIFF and applied SR which put it at 12000 x 8000. I also exported it as a Jpeg @1600 on the long side. Should the file size not be twice as big? I also created a virtual copy of the original RAW file and applied SR to it. Same results.

 

[Conrad Chavez]

The file size of a JPEG doesn’t depend only on the width and height in pixels, because of the compression that JPEG uses. If the image content has large areas that are more easily compressible by the lossy JPEG algorithm (for example, small color variations, low amount of detail, or low noise), the file size savings can be out of proportion to the difference in the number of pixels.

 

[Johan Elzenga]

I also exported it as a Jpeg @1600 on the long side. Should the file size not be twice as big?

Not if you resize the image on export…

 

[Zenon]

Thanks everyone. So am I getting any benefit from using SR? I would normally not apply this to the entire file. It would be for a cropping wildlife shots. Doing tests with crops sometimes it’s hard to tell if I got more detail by applying it. Or enough to make it worth it.

I’m trying to better understand the purpose of upscaling.

 

[clee01l]

Thanks everyone. So am I getting any benefit from using SR?

Super Resolution has nothing to do with size. It does have something to do with resolution.

Here's how I see it. If you take 4 pixels in a square and apply SR you end up with 16 pixels in a square. The individual pixel in the 16 pixel square is a blend (smart blend?) of 9 adjacent pixels. If you have ever blown up a standard image to some maximum, you start to see pixelation. While an image with SR applied has 4 times the number of original pixels, when blown up to the same size, the pixelation won't be (as) apparent. Super Resolution is very different than upsize on Export using the resize process.

SR is done on the whole image prior to cropping because the edge pixels won't have adjacent pixels on the edge to contribute to the analysis. Where as a crop after SR will have edge pixels that are a blend of adjacent pixels outside of the crop window.

 

[Zenon]

Super Resolution has nothing to do with size. It does have something to do with resolution.

Here's how I see it. If you take 4 pixels in a square and apply SR you end up with 16 pixels in a square. The individual pixel in the 16 pixel square is a blend (smart blend?) of 9 adjacent pixels. If you have ever blown up a standard image to some maximum, you start to see pixelation. While an image with SR applied has 4 times the number of original pixels, when blown up to the same size, the pixelation won't be (as) apparent. Super Resolution is very different than upsize on Export using the resize process.

SR is done on the whole image prior to cropping because the edge pixels won't have adjacent pixels on the edge to contribute to the analysis. Where as a crop after SR will have edge pixels that are a blend of adjacent pixels outside of the crop window.

Thanks for explanation.

 

[Gnits]

Good explanation.

 

[Johan Elzenga]

Thanks everyone. So am I getting any benefit from using SR? I would normally not apply this to the entire file. It would be for a cropping wildlife shots. Doing tests with crops sometimes it’s hard to tell if I got more detail by applying it. Or enough to make it worth it.

I’m trying to better understand the purpose of upscaling.

Super resolution is meant for situations where you want to make a large print, and the number of pixels in your image is too low for that. As a result, the print resolution would become too low and so you’ll see individual pixels (jagged edges). You are doing the exact opposite, however. You downsized the exported images to 1600 pixels at the longest side (unless I misunderstood that). That means you’ll simply undo any effect that super resolution has, because in both cases you end up with an image that is only 1600 pixels wide, whereas the original was 6000 pixels wide. There is indeed no benefit in downsizing from 12000 pixels instead of 6000 pixels, if those 12000 pixels were produced by upsizing the 6000 pixel image…

If you cropped the image to 1600 pixels wide, then of course it is a different story.

 

[Zenon]

Super resolution is meant for situations where you want to make a large print, and the number of pixels in your image is too low for that. As a result, the print resolution would become too low and so you’ll see individual pixels (jagged edges). You are doing the exact opposite, however. You downsized the exported images to 1600 pixels at the longest side (unless I misunderstood that). That means you’ll simply undo any effect that super resolution has, because in both cases you end up with an image that is only 1600 pixels wide, whereas the original was 6000 pixels wide. There is indeed no benefit in downsizing from 12000 pixels instead of 6000 pixels, if those 12000 pixels were produced by upsizing the 6000 pixel image…

If you cropped the image to 1600 pixels wide, then of course it is a different story.

It's 1600 pixels on the longest side for posting on websites. I was wondering if it would help out wth some of the crops I push to the edge. By that I mean as much as I can while maintaining detail. I guess a lot of people must print really big prints for Topaz to offer that as a stand alone app. I guess there is a market for that requirement.

Thanks.

 

[Johan Elzenga]

Yes, it will. Do the math. Your original image is 6000 pixels wide, but if your crop leaves a resulting image that is smaller than 1600 pixels wide, then super resolution can help. Let's take a practical example with some numbers to make it clear. Suppose you shot a bird in flight, but the bird was a quite far away. You crop the image so you have a nice bird picture, but that picture is now only 1000 pixels wide. If you would place that on your website, you would have to upscale it from 1000 pixels to 1600 pixels. If you use super resolution, then that same crop will be 2000 pixels wide, so now you can downscale it to 1600 pixels. Assuming that super resolution is better than 'normal' upscaling, this will give a better result.

 

[Zenon]

Yes, it will. Do the math. Your original image is 6000 pixels wide, but if your crop leaves a resulting image that is smaller than 1600 pixels wide, then super resolution can help. Let's take a practical example with some numbers to make it clear. Suppose you shot a bird in flight, but the bird was a quite far away. You crop the image so you have a nice bird picture, but that picture is now only 1000 pixels wide. If you would place that on your website, you would have to upscale it from 1000 pixels to 1600 pixels. If you use super resolution, then that same crop will be 2000 pixels wide, so now you can downscale it to 1600 pixels. Assuming that super resolution is better than 'normal' upscaling, this will give a better result.

Thanks. Even before you stated this just from random tests it appeared there was that tipping point between where it makes a difference to apply. 1000 is a pretty hefty crop and for wildlife. I rarely crop more than 50% for that.

 

[Zenon]

You can really see the difference when cropping to 1000, apply SR and then exporting at 1600. Not very practical for me. Occasionally at times I do export larger files. Lets say I crop to 2800 x 1863 and what to export at 3600. Now I see where I get the advantage of SR. Since I always avoided do that for obvious reasons I now have another tool. Glad I looked into it.

Thanks again everyone.

 

[Zenon]

I decided to try it out. I could not get closer and when I did approach a bit it flew off. I cropped below export size, applied it and exported. I would not print or put it on the wall but it's OK for a few casual photo sharing forums.