Sony Extended Zoom, Images Here (not words!)

I don't have much need to upsize images, so I was looking at other Topaz AI modules, such as the new Sharpen AI -- have you used any of them?
Yes, I was wondering why you were interested in AI Gigapixel.

No I don't use any other Topaz AI modules just Topaz DeNoise but that's ancient and not AI.

This thread may be of interested to you

https://www.dpreview.com/forums/post/62443662

David
I've downloaded Sharpen AI, and tried it out on the Great Egret Shot. I notice that it loses the exif data -- does Gigapixel do that too?
There is an option to save metadata in AI Gigapixel.
Yes, I found and used it, as you can see below. There seems to be no such option in Sharpen AI. Obviously, there are complicated ways of getting round this, but, like you, it would put me off buying it. It's a very new program, so perhaps they'll rectify the omission.
I very much doubt whether I would have bought the program if this wasn't possible.

aa84c58ba2d54537a547092fb2749075.jpg
Anyway, here's the result:

Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results
Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results

For comparison, here's the original again:

Original file, as output from DxO PL2
Original file, as output from DxO PL2
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
PL2 has excellent sharpening to compensate for lens flaws, plus USM, but DxO advises against using USM.
 
Yup, very nice indeed,



a543798fd59e4c789291b1594ed5c78d.jpg.png



--
Elliott
 
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
BTW, you didn't comment on the Gigapixel version I showed. I did try resizing the doubled version back to the original size, but it didn't change anything from the original.
 
Anyway, here's the result:

Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results
Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results

And here's a 2x upscale using AI Gigapixel:

Dimensions doubled in each dimensions (ie, what CIZ might do)
Dimensions doubled in each dimensions (ie, what CIZ might do)

For comparison, here's the original again:

Original file, as output from DxO PL2
Original file, as output from DxO PL2
Nigel,

This is very interesting, a few comments.

First, Except for the detail in the feet, very minimally sharpened, I cannot see any meaningful difference between the 3 images. I downloaded them, viewed them large, I had to read captions to tell which version I was viewing. Took me a while to find the slight difference in the feet.

The last up-scaled version, from 5.8mp to 4 times, 23.5mp, looks no different at full screen size. The difference would be printing large or viewing on a large TV.

The up-scaling, not only un-apparent, has not changed the color, exposure, it changed 'Nothing' as Shultz used to say. That's some very good pixel invention, 75% are digitally guessed at and created.

........................

Math, in round numbers:

This is a heavy crop, it is only 5.8mp, from a 20mp original, i.e. kind of like using SZ at 5mp, which with your RX10m4 would be SZ 2.0x = 600mm x 2 = equiv of 1,200mm. So you made a post crop which is technically the same as an SZ 2.0x in-camera optical crop.

For the best comparative Jpeg, you would need to change to Jpeg Only, Extra Fine Compression, shoot OZ 20mp.

Academically, your up-scaling software/your computer, vs Sony in-camera up-scaling could be compared if you had also taken a separate Jpeg, extra fine shot with CIZ on, then keep zooming to get CIZ 2.0x 1,200mm, which CIZ would upscale back to 20mp.

It would convert the idea of post upscaling using a powerful computer vs in-camera into Results, not Words.

And, my problem with viewing RAW images, is that any RAW image already has variable PP by selected RAW software and that photographer's settings/choices/preferences, variable sharpening, variable noise Reduction, .... so I don't know what I am looking at. Am I looking at something I could do? No. Will I learn RAW? No. Just give up now!

I do know, if good PP skills are involved, RAW can beat Jpeg, which is why RAW shooters would never use SZ or CIZ. This is a point you don't concede, SZ and/or CIZ is for the Jpeg Crowd, my people!

--
Elliott
 
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
BTW, you didn't comment on the Gigapixel version I showed. I did try resizing the doubled version back to the original size, but it didn't change anything from the original.
Unless I'm going potty, there was no Gigapixel version in the post I replied to. I can see it now though.

Will reply again to that post.

Doh!

David
 
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
BTW, you didn't comment on the Gigapixel version I showed. I did try resizing the doubled version back to the original size, but it didn't change anything from the original.
Unless I'm going potty, there was no Gigapixel version in the post I replied to. I can see it now though.

Will reply again to that post.
I confess, I edited the post shortly after first posting.

I only decided to download and try out Gigapixel after doing the original post. Luckily, it didn't take long to download and install, there's nothing to learn, and it runs very fast. So I decided to add that image to my post, rather than doing a new post.
 
My conclusion: Digital zoom combined with digital works surprisingly well for static objects, however for animals the results are blurry and with reduced contrast.

14X zoom equivalent
14X zoom equivalent

12X
12X

2b9ce8468f934427948dfc251788ddcb.jpg
 
Last edited:
Anyway, here's the result:

Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results
Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results

And here's a 2x upscale using AI Gigapixel:

Dimensions doubled in each dimensions (ie, what CIZ might do)
Dimensions doubled in each dimensions (ie, what CIZ might do)

For comparison, here's the original again:

Original file, as output from DxO PL2
Original file, as output from DxO PL2
Nigel,

This is very interesting, a few comments.

First, Except for the detail in the feet, very minimally sharpened, I cannot see any meaningful difference between the 3 images. I downloaded them, viewed them large, I had to read captions to tell which version I was viewing. Took me a while to find the slight difference in the feet.

The last up-scaled version, from 5.8mp to 4 times, 23.5mp, looks no different at full screen size. The difference would be printing large or viewing on a large TV.

The up-scaling, not only un-apparent, has not changed the color, exposure, it changed 'Nothing' as Shultz used to say. That's some very good pixel invention, 75% are digitally guessed at and created.

........................

Math, in round numbers:

This is a heavy crop, it is only 5.8mp, from a 20mp original, i.e. kind of like using SZ at 5mp, which with your RX10m4 would be SZ 2.0x = 600mm x 2 = equiv of 1,200mm. So you made a post crop which is technically the same as an SZ 2.0x in-camera optical crop.
Agreed, but I wouldn't have been able to get the right crop in-camera. I was shooting from some distance away, on a moving boat. Many of my other shots, even using the full sensor, missed out the bird's feet. And in none of them was the bird centred. It was much better to crop in post.
For the best comparative Jpeg, you would need to change to Jpeg Only, Extra Fine Compression, shoot OZ 20mp.

Academically, your up-scaling software/your computer, vs Sony in-camera up-scaling could be compared if you had also taken a separate Jpeg, extra fine shot with CIZ on, then keep zooming to get CIZ 2.0x 1,200mm, which CIZ would upscale back to 20mp.

It would convert the idea of post upscaling using a powerful computer vs in-camera into Results, not Words.
Yes, when I get a chance, I'll shoot an image on a tripod, first in RAW+Extra fine JPEG, and then CIZ JPEG. I'll upscale the RAW 2x, and compare it to the CIZ OOC result.
And, my problem with viewing RAW images, is that any RAW image already has variable PP by selected RAW software and that photographer's settings/choices/preferences, variable sharpening, variable noise Reduction, .... so I don't know what I am looking at. Am I looking at something I could do? No. Will I learn RAW? No. Just give up now!

I do know, if good PP skills are involved, RAW can beat Jpeg, which is why RAW shooters would never use SZ or CIZ. This is a point you don't concede, SZ and/or CIZ is for the Jpeg Crowd, my people!
I do understand that many people shoot only JPEGS, and aren't fond of post-processing. I also understand that cropping is useful, and do it all the time myself, in most images (not usually as heavily as this image). But what you haven't explained are two things:
  • Why do you need to upscale the in-camera cropped images? Surely it would be better to just leave them as 5 or 10mp crops? The files are smaller, you can fit more on a card and in the buffer, continuous shooting works much better, they're the right size for uploading to the web or emailing, etc? Only in the special case of needing to print very large, high res images is upscaling needed, and that's best done in post.
  • Why not just crop in post-processing, rather than in-camera? You usually seem to further crop the in-camera cropped images anyway, so why not just shoot full size, and crop as needed later? Yes, I know you like the magnified view when shooting, but you can get that anyway, and with better control of focus and exposure, including expanded flexible spot and spot exposure.
 
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
BTW, you didn't comment on the Gigapixel version I showed. I did try resizing the doubled version back to the original size, but it didn't change anything from the original.
Unless I'm going potty, there was no Gigapixel version in the post I replied to. I can see it now though.

Will reply again to that post.
I confess, I edited the post shortly after first posting.

I only decided to download and try out Gigapixel after doing the original post. Luckily, it didn't take long to download and install, there's nothing to learn, and it runs very fast. So I decided to add that image to my post, rather than doing a new post.
I've just tried scal1ing up a full size 20mp image using Gigapixel: it took just over two minutes. I don't know how that compares with your machine? It's a bit slow if you're sitting watching, but no problem for a batch job. Here's the original image:

Original, full-size, uncropped image (16Mb)
Original, full-size, uncropped image (16Mb)

View attachment e3aa8e2b8f894368a308d3c2ed7a1f99.jpg
2x upscaled version (53.5Mb), processing took about 125 seconds

And here's one that's had the Sharpen AI treatment. The raccoon's face was already pretty sharp, so the changes are subtle, and it doesn't seem to have messed up out-of-focus areas:

Sharpen AI applied to full-size image
Sharpen AI applied to full-size image

I don't think this image really needed the sharpening, but I've seen others where the benefit is much greater.
 
Last edited:
I don't have much need to upsize images, so I was looking at other Topaz AI modules, such as the new Sharpen AI -- have you used any of them?
Yes, I was wondering why you were interested in AI Gigapixel.

No I don't use any other Topaz AI modules just Topaz DeNoise but that's ancient and not AI.

This thread may be of interested to you

https://www.dpreview.com/forums/post/62443662

David
I've downloaded Sharpen AI, and tried it out on the Great Egret Shot. I notice that it loses the exif data .

Anyway, here's the result:

Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results
Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results

And here's a 2x upscale using AI Gigapixel:

Dimensions doubled in each dimensions (ie, what CIZ might do)
Dimensions doubled in each dimensions (ie, what CIZ might do)

For comparison, here's the original again:

Original file, as output from DxO PL2
Original file, as output from DxO PL2
Something fishy going on here, I swear there was only two instances of this image last time I replied!

This is an example of needing a good image to start with. The beak eye and head of the bird are not sharp or in focus. It's one or the other or both. It also has blown highlights and lacks detail. So that area made double the size by Megapixel will look exactly twice as bad as the original!

Here's the original 100% crop and like I said it's pretty bad.

original
original

gigapixel
gigapixel

Viewed in the post, the image quality or lack of will look identical but view them at 100% and the Gigapixel image will look exactly twice as bad because you doubled the size.

But if the head eye and beak had been pin sharp to start with, they would still look pin sharp after being through Gigapixel even at 100% and double the size of the original.

In the latter scenario, the Gigapixel image wouldn't look twice as good as the original, it would look equally as good but twice as big. You cannot say that about CIZ because for a start it will will add a raft of jpeg artifacts to the image. It's the first thing I spot when folk post photos of birds at 2x CIZ.

None of what I said really matters if you don't pixel peep. :-)

David
 
Last edited:
I don't have much need to upsize images, so I was looking at other Topaz AI modules, such as the new Sharpen AI -- have you used any of them?
Yes, I was wondering why you were interested in AI Gigapixel.

No I don't use any other Topaz AI modules just Topaz DeNoise but that's ancient and not AI.

This thread may be of interested to you

https://www.dpreview.com/forums/post/62443662

David
I've downloaded Sharpen AI, and tried it out on the Great Egret Shot. I notice that it loses the exif data .

Anyway, here's the result:

Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results
Processed in Topaz AI, using the Sharpen mode, which seemed to produce the most realistic results

And here's a 2x upscale using AI Gigapixel:

Dimensions doubled in each dimensions (ie, what CIZ might do)
Dimensions doubled in each dimensions (ie, what CIZ might do)

For comparison, here's the original again:

Original file, as output from DxO PL2
Original file, as output from DxO PL2
Something fishy going on here, I swear there was only two instances of this image last time I replied!

This is an example of needing a good image to start with. The beak eye and head of the bird are not sharp or in focus. It's one or the other or both. It also has blown highlights and lacks detail. So that area made double the size by Megapixel will look exactly twice as bad as the original!
Yes, I was hoping it would magically improve the image, but it didn't.
Here's the original 100% crop and like I said it's pretty bad.

original
original

gigapixel
gigapixel

Viewed in the post, the image quality or lack of will look identical but view them at 100% and the Gigapixel image will look exactly twice as bad because you doubled the size.

But if the head eye and beak had been pin sharp to start with, they would still look pin sharp after being through Gigapixel even at 100% and double the size of the original.

In the latter scenario, the Gigapixel image wouldn't look twice as good as the original, it would look equally as good but twice as big. You cannot say that about CIZ because for a start it will will add a raft of jpeg artifacts to the image. It's the first thing I spot when folk post photos of birds at 2x CIZ.

None of what I said really matters if you don't pixel peep. :-)

David
I've just tried another test with the raccoon shot.

I upscaled the 100% version using Gigapixel and also using an average of multiple algorithms in PSP (quick and easy). It's a sharp image, and Gigapixel did a good job maintaining quality at 200%. But I really had to pixel peep, viewing both images at 100% side-by-side, to see much difference, and even then it was slight.

Given how rarely I need to upscale an image, and how slight the improvement is in Gigapixel compared to the fast rescaling algorithm in PSP, I don't think I'll bother with it.

I think I could make more use of Sharpen AI, but it's a pain that the current early release doesn't retain the metadata. Obviously there are workarounds, but they're inconvenient.
 
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
BTW, you didn't comment on the Gigapixel version I showed. I did try resizing the doubled version back to the original size, but it didn't change anything from the original.
Unless I'm going potty, there was no Gigapixel version in the post I replied to. I can see it now though.

Will reply again to that post.
I confess, I edited the post shortly after first posting.

I only decided to download and try out Gigapixel after doing the original post. Luckily, it didn't take long to download and install, there's nothing to learn, and it runs very fast. So I decided to add that image to my post, rather than doing a new post.
I've just tried scal1ing up a full size 20mp image using Gigapixel: it took just over two minutes. I don't know how that compares with your machine? It's a bit slow if you're sitting watching, but no problem for a batch job. Here's the original image:
I wouldn't know, Nigel. I wouldn't dream of scaling up a full size 20mp image. I use Gigapixel to upscale 2 or to 4mp crops to 8mp and they take around a minute so I would image a 20mp image would take a lot longer than two minutes more like 10 minutes but I'll try one later and let you know.
Original, full-size, uncropped image (16Mb)
Original, full-size, uncropped image (16Mb)

View attachment e3aa8e2b8f894368a308d3c2ed7a1f99.jpg
2x upscaled version (53.5Mb), processing took about 125 seconds

And here's one that's had the Sharpen AI treatment. The raccoon's face was already pretty sharp, so the changes are subtle, and it doesn't seem to have messed up out-of-focus areas:

Sharpen AI applied to full-size image
Sharpen AI applied to full-size image

I don't think this image really needed the sharpening, but I've seen others where the benefit is much greater.
This is much better than your Great Egret effort. But honestly I'm not sure why you want such a huge image. To me it was big enough to start with! I would have done the opposite and downsized it!

Obviously AI Gigapixel can upscale huge images but the demonstrations I've seen on the Topaz website revolve round increasing the size of small to medium sized images.

David
 
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
BTW, you didn't comment on the Gigapixel version I showed. I did try resizing the doubled version back to the original size, but it didn't change anything from the original.
Unless I'm going potty, there was no Gigapixel version in the post I replied to. I can see it now though.

Will reply again to that post.
I confess, I edited the post shortly after first posting.

I only decided to download and try out Gigapixel after doing the original post. Luckily, it didn't take long to download and install, there's nothing to learn, and it runs very fast. So I decided to add that image to my post, rather than doing a new post.
Thank god for that. I wasn't going to have a drink to tonight, thinking the ale was pickling my brain but now I'm going to have an extra one!

David
 
I do understand that many people shoot only JPEGS, and aren't fond of post-processing. I also understand that cropping is useful, and do it all the time myself, in most images (not usually as heavily as this image). But what you haven't explained are two things:
  • Why do you need to upscale the in-camera cropped images? Surely it would be better to just leave them as 5 or 10mp crops? The files are smaller, you can fit more on a card and in the buffer, continuous shooting works much better, they're the right size for uploading to the web or emailing, etc? Only in the special case of needing to print very large, high res images is upscaling needed, and that's best done in post.
'super magnification'

Keep in mind, I am mostly in S mode shooting motion. And stills or fixed zone distance motion like wrestling, or pre-focused on the goalie, I toggle AF/MF frequently, using focus magnification/peaking often. Sometimes MF just to verify focus, then view over the camera to time the shot. I'm shooting thru dirty plexiglass, nets, fences, ....

so I view via 'super magnification' a lot and I have gotten very quick using it.

..................................................

In the past I have been using and showing primarily CIZ 2.0x which crops and up-scales back to 20mp. rx100m1 CIZ to 200mm; rx100m3 CIZ to 140mm.

You are right, and proven by your 3 photos, up-scaling is not needed at typical viewing size, no visual difference, unless printing large. But, keep in mind, CIZ gives more extended zoom than SZ does, for that reason alone someone might use/prefer CIZ.

.........................................

Implementation:

CIZ + SZ or SZ only via ST. ST Toggle is what has me using only SZ now.

My Oly has 2X which I use without hesitation. It is like Sony's CIZ, Crop and up-scale. I think it's upscale is slightly better than Sony, but certainly nothing obvious. Turns 28-300mm OZ (constant f2.8 lens) into 28-600mm 2X, still constant f2.8. As ALL Sony SZ or CIZ occurs after the lens is at it's darkest, that is probably the difference, unless lots of light.

Previously I much preferred Oly's Implementation.

No setup, no menu anything. It is ALWAYS available via a dedicated button. It is instant doubling and instant back to OZ. It doubles any view you have, zoom to 235 OZ, press 2x, jumps to 470mm. After that jump, you can zoom in to the max 600mm 2x or out to 56mm (28 2X). Press again instantly back to where you were, or, if you moved, 1/2 of whatever.

rx100m1 did not have ST smart telecon, so I used both SZ and CIZ. The rx100m6 has ST Smart Telecon Toggle. Using that, is even better than Oly. AFTER you set it up (a confusing search to change 2 things in menu), it is a 2 step toggle: OZ;1.4x/2.0x/OZ. Using ST Toggle, CIZ is blocked, so no CIZ for now.

I am getting ready for SZ 280mm using 10mp, for Ben's soccer about to start. Watch until he gets the ball, presto, instant close up of his vicinity. Where'd he go? Presto, back to wider view, find him, press again when ...

I only want a few shots. I like the challenge of getting a single shot, not a spray of continuous. I hate editing continuous, so my methods give me the magnified view to time the shot. Also, using MF, finishing a pre-focused shot is a speck faster than AF.

After that, I will be trying 20mp CIZ, progressive (slow) zoom, 200OZ to 400mm CIZ.
  • Why not just crop in post-processing, rather than in-camera? You usually seem to further crop the in-camera cropped images anyway, so why not just shoot full size, and crop as needed later? Yes, I know you like the magnified view when shooting, but you can get that anyway, and with better control of focus and exposure, including expanded flexible spot and spot exposure.
Without a doubt, I see and time the shots better while zoomed, and I discover things/details of far off subjects I might have missed restricted to 200mm. Full Area Focus and Metering seems like a limitation, but, you are so zoomed in, it is like using medium or large spot area. I also use AEL frequently, whether OZ or Extended Zoom, and often too I am using MF.

..................................

Aside from how I use it, for others, they should know that a few small sensor extra reach cameras lurk inside of their pocket rocket. I love your conceptualization.

Oh yeah, most small sensor zooms lenses have darker lenses both to start and finish. All their extra reach is at their darker end.

Sony Extended Zooms occur at brighter apertures than small sensor zooms.

................................

Extend Reach while Maintaining Lens Brightness. (SZ via ST)

note: rx100m6 has already reached f4.0 at 40mm. It narrows to f4.5 at 110mm


I just discovered this. ST Toggle doubles wherever you are, at the aperture are using not just past optical, so:

start: 24mm f2.8, 20mp

press to 1.4x = 34mm still f2.8 10mp

press to 2.0x = 48mm still f2.8, 5mp (OZ would be f4.0)

zoom a bit, say

start: 35mm f3.5.

press to 1.4x = 50mm, still f3.5, 10mp

press to 2.0x = 70mm, still f3.5, 5mp

repeat: rx100m6 is already f4.0 at 40mm. It narrows to f4.5 at 110mm
 
Last edited:
Extend Reach while Maintaining Lens Brightness. (SZ via ST)

note: rx100m6 has already reached f4.0 at 40mm. It narrows to f4.5 at 110mm

I just discovered this. ST Toggle doubles wherever you are, at the aperture are using not just past optical, so:

start: 24mm f2.8, 20mp

press to 1.4x = 34mm still f2.8 10mp

press to 2.0x = 48mm still f2.8, 5mp (OZ would be f4.0, 20mp)

zoom a bit, say

start: 35mm f3.5.

press to 1.4x = 50mm, still f3.5, 10mp

press to 2.0x = 70mm, still f3.5, 5mp

repeat: rx100m6 is already f4.0 at 40mm. It narrows to f4.5 at 110mm

--
Elliott
 
Last edited:
I can see the AI image is sharper but that's all I can say really. I wouldn't need a sharpening module because I swear by USM (Unsharp Mask) in Photoshop. Perhaps it's because I've been using it for 20 years and never felt the need to try anything different. I don't even like Lightroom's attempt at sharpening.
BTW, you didn't comment on the Gigapixel version I showed. I did try resizing the doubled version back to the original size, but it didn't change anything from the original.
Unless I'm going potty, there was no Gigapixel version in the post I replied to. I can see it now though.

Will reply again to that post.
I confess, I edited the post shortly after first posting.

I only decided to download and try out Gigapixel after doing the original post. Luckily, it didn't take long to download and install, there's nothing to learn, and it runs very fast. So I decided to add that image to my post, rather than doing a new post.
I've just tried scal1ing up a full size 20mp image using Gigapixel: it took just over two minutes. I don't know how that compares with your machine? It's a bit slow if you're sitting watching, but no problem for a batch job. Here's the original image:
I wouldn't know, Nigel. I wouldn't dream of scaling up a full size 20mp image. I use Gigapixel to upscale 2 or to 4mp crops to 8mp and they take around a minute so I would image a 20mp image would take a lot longer than two minutes more like 10 minutes but I'll try one later and let you know.
I tried a 1220x915 crop, scaled up by a factor of four. It took 10 secs.
Original, full-size, uncropped image (16Mb)
Original, full-size, uncropped image (16Mb)

View attachment e3aa8e2b8f894368a308d3c2ed7a1f99.jpg
2x upscaled version (53.5Mb), processing took about 125 seconds

And here's one that's had the Sharpen AI treatment. The raccoon's face was already pretty sharp, so the changes are subtle, and it doesn't seem to have messed up out-of-focus areas:

Sharpen AI applied to full-size image
Sharpen AI applied to full-size image

I don't think this image really needed the sharpening, but I've seen others where the benefit is much greater.
This is much better than your Great Egret effort. But honestly I'm not sure why you want such a huge image. To me it was big enough to start with! I would have done the opposite and downsized it!
I didn't, it was just a performance test.
Obviously AI Gigapixel can upscale huge images but the demonstrations I've seen on the Topaz website revolve round increasing the size of small to medium sized images.
Yup, much more useful.
 
You are right, and proven by your 3 photos, up-scaling is not needed at typical viewing size, no visual difference, unless printing large. But, keep in mind, CIZ gives more extended zoom than SZ does, for that reason alone someone might use/prefer CIZ.
Does it? I thought they were both 5mp crops at maximum 'zoom'.
 
Extend Reach while Maintaining Lens Brightness. (SZ via ST)

note: rx100m6 has already reached f4.0 at 40mm. It narrows to f4.5 at 110mm

I just discovered this. ST Toggle doubles wherever you are, at the aperture are using not just past optical, so:

start: 24mm f2.8, 20mp

press to 1.4x = 34mm still f2.8 10mp

press to 2.0x = 48mm still f2.8, 5mp (OZ would be f4.0, 20mp)

zoom a bit, say

start: 35mm f3.5.

press to 1.4x = 50mm, still f3.5, 10mp

press to 2.0x = 70mm, still f3.5, 5mp

repeat: rx100m6 is already f4.0 at 40mm. It narrows to f4.5 at 110mm
Your deluding yourself. Elliott. I could say the exact same thing when cropping in post but it would be meaningless. You haven't got 1.4x or 2x anything. The only thing you've got is less pixels than if you'd shot at full image size. You haven't gained any extra light either. Let's say you shot at 24mm F2.8 1/250 and ISO 200 with SZ 5mp crop. You say that would equal 48mm focal length. But if you zoom to 48mm with the sensor at its native resolution both images would cover the exact same field of view. The big difference is the optical zoomed image would have 4x the number of pixels. Granted because the lens is now down to F4, if you wanted the same shutter speed of 1/250 you'd have to shoot at 400 ISO. Would you get a nosier image? No, not at all, at the same viewing size they'd be exactly the same. To view them at the same size you'd have to downsize the full res image and in doing so it reduces the noise to the same level as your 5mb crop at 200 ISO. There's no such thing as a free lunch as far as aperture is concerned.

David
 
Extend Reach while Maintaining Lens Brightness. (SZ via ST)

note: rx100m6 has already reached f4.0 at 40mm. It narrows to f4.5 at 110mm

I just discovered this. ST Toggle doubles wherever you are, at the aperture are using not just past optical, so:

start: 24mm f2.8, 20mp

press to 1.4x = 34mm still f2.8 10mp

press to 2.0x = 48mm still f2.8, 5mp (OZ would be f4.0, 20mp)

zoom a bit, say

start: 35mm f3.5.

press to 1.4x = 50mm, still f3.5, 10mp

press to 2.0x = 70mm, still f3.5, 5mp

repeat: rx100m6 is already f4.0 at 40mm. It narrows to f4.5 at 110mm
Your deluding yourself. Elliott. I could say the exact same thing when cropping in post but it would be meaningless. You haven't got 1.4x or 2x anything. The only thing you've got is less pixels than if you'd shot at full image size. You haven't gained any extra light either. Let's say you shot at 24mm F2.8 1/250 and ISO 200 with SZ 5mp crop. You say that would equal 48mm focal length. But if you zoom to 48mm with the sensor at its native resolution both images would cover the exact same field of view. The big difference is the optical zoomed image would have 4x the number of pixels. Granted because the lens is now down to F4, if you wanted the same shutter speed of 1/250 you'd have to shoot at 400 ISO. Would you get a nosier image? No, not at all, at the same viewing size they'd be exactly the same. To view them at the same size you'd have to downsize the full res image and in doing so it reduces the noise to the same level as your 5mb crop at 200 ISO. There's no such thing as a free lunch as far as aperture is concerned.

David
I carefully added the MP count in the examples above. ANY SZ is the same as a post crop, no difference whatsoever, except, you can see a bit better when shooting, the crop is effortless, and, as we think, using OZ to get close, say 40mm, the aperture would be f4.0.

IF I needed a bit faster shutter, or IF I wanted a bit less noise, using SZ (via ST only). I could stay wider for wider aperture, and then get my EFFORTLESS in-camera crop.

Seeing large, timing the shot, effortless, that's all it is.


Using SZ via ST, you can stay wide view and aperture, and toggle closer view, lowering the mp (as a post crop would) as you toggle closer.

You can never get any CIZ except at f4.5 with rx100m6, then, all the extra reach (like a post crop) remains f4.5, then, the camera up-scales it for you.

CIZ gives more extended reach than SZ. Go back and read the other thread, ... 'Explanation'
 
Extend Reach while Maintaining Lens Brightness. (SZ via ST)

note: rx100m6 has already reached f4.0 at 40mm. It narrows to f4.5 at 110mm

I just discovered this. ST Toggle doubles wherever you are, at the aperture are using not just past optical, so:

start: 24mm f2.8, 20mp

press to 1.4x = 34mm still f2.8 10mp

press to 2.0x = 48mm still f2.8, 5mp (OZ would be f4.0, 20mp)

zoom a bit, say

start: 35mm f3.5.

press to 1.4x = 50mm, still f3.5, 10mp

press to 2.0x = 70mm, still f3.5, 5mp

repeat: rx100m6 is already f4.0 at 40mm. It narrows to f4.5 at 110mm
Your deluding yourself. Elliott. I could say the exact same thing when cropping in post but it would be meaningless. You haven't got 1.4x or 2x anything. The only thing you've got is less pixels than if you'd shot at full image size. You haven't gained any extra light either. Let's say you shot at 24mm F2.8 1/250 and ISO 200 with SZ 5mp crop. You say that would equal 48mm focal length. But if you zoom to 48mm with the sensor at its native resolution both images would cover the exact same field of view. The big difference is the optical zoomed image would have 4x the number of pixels. Granted because the lens is now down to F4, if you wanted the same shutter speed of 1/250 you'd have to shoot at 400 ISO. Would you get a nosier image? No, not at all, at the same viewing size they'd be exactly the same. To view them at the same size you'd have to downsize the full res image and in doing so it reduces the noise to the same level as your 5mb crop at 200 ISO. There's no such thing as a free lunch as far as aperture is concerned.

David
I carefully added the MP count in the examples above. ANY SZ is the same as a post crop, no difference whatsoever, except, you can see a bit better when shooting, the crop is effortless, and, as we think, using OZ to get close, say 40mm, the aperture would be f4.0.

IF I needed a bit faster shutter, or IF I wanted a bit less noise, using SZ (via ST only). I could stay wider for wider aperture, and then get my EFFORTLESS in-camera crop.

Seeing large, timing the shot, effortless, that's all it is.


Using SZ via ST, you can stay wide view and aperture, and toggle closer view, lowering the mp (as a post crop would) as you toggle closer.

You can never get any CIZ except at f4.5 with rx100m6, then, all the extra reach (like a post crop) remains f4.5, then, the camera up-scales it for you.

CIZ gives more extended reach than SZ. Go back and read the other thread, ... 'Explanation'
How does CIZ give more reach than SZ? At maximum, they're both 5mp centre crops of 72mm focal length images.
 

Keyboard shortcuts

Back
Top