Larry H-L
Well-known
I may have missed a post about it here, but if it hasn't been mentioned, Adobe introduced a new feature in Photoshop's Adobe Camera Raw v 13.2. The new addition is "high-resolution mode" paired with the previous "enhance details" function.
This new mode takes RAW, TIF, or JPEG files and boosts the resolution by 4x. A 6mp RD-1 suddenly becomes capable of delivering 24mp files. The new software supposedly uses machine learning to build the files.
I tried a couple of 24mp Sony RAW files and compared the details in the original and in the high rez version. I couldn't see any loss of detail in the new file. It almost seems too good to be true. The only downside is that it is relatively slow to build each file, but worth it if you are planning on printing large sizes.
Has anyone tried it?
This new mode takes RAW, TIF, or JPEG files and boosts the resolution by 4x. A 6mp RD-1 suddenly becomes capable of delivering 24mp files. The new software supposedly uses machine learning to build the files.
I tried a couple of 24mp Sony RAW files and compared the details in the original and in the high rez version. I couldn't see any loss of detail in the new file. It almost seems too good to be true. The only downside is that it is relatively slow to build each file, but worth it if you are planning on printing large sizes.
Has anyone tried it?
Larry H-L
Well-known
I should say that resolution is doubled, resulting in a 4x larger file.
Here is Adobe Evangelista Julieanne Kost explaining the feature:
https://www.youtube.com/watch?v=-uNdijku7iY
Here is Adobe Evangelista Julieanne Kost explaining the feature:
https://www.youtube.com/watch?v=-uNdijku7iY
shawn
Veteran
I've run some Epson files through Topaz Gigapixel AI and the results can be impressive.
Color differences are just from processing differences between the files.
Shawn
Color differences are just from processing differences between the files.
Shawn
lynnb
Veteran
My first reaction on seeing this news was to think this could be a lifeline for micro 4/3. I can imagine there's a lot of smiling RD-1 owners now.
shawn
Veteran
Topaz Sharpen AI can also bring up detail... combining the two can be interesting.
https://www.dpreview.com/forums/post/64471184
https://www.dpreview.com/forums/post/64471203
Shawn
https://www.dpreview.com/forums/post/64471184
https://www.dpreview.com/forums/post/64471203
Shawn
Larry H-L
Well-known
That's impressive Shawn.
These types of software enhancements appear to work really well, and it certainly seems like a decent alternative to constantly buying camera body upgrades and bags of new lenses.
Tomorrow I am going to try it on some RD-1 and Merrill files to see how it looks.
These types of software enhancements appear to work really well, and it certainly seems like a decent alternative to constantly buying camera body upgrades and bags of new lenses.
Tomorrow I am going to try it on some RD-1 and Merrill files to see how it looks.
willie_901
Veteran
There Are No Miracles
There Are No Miracles
But that doesn't mean one shouldn't use methods that make the most out of their data.
I don't think Adobe is doing anything new or unique.
What Adobe has done is increase the feature set of Camera Raw. Resolution is not increased. A mathematical model is used to estimate the information content of pixels that never existed. There is no new information in the modeled image. Images created by these methods don't have more information content nor do they the same information content as a higher resolution sensor image.
Adobe, and others, are taking advantage of higher performance computers to render images with superior perceived image quality.
This does not the model's parameter estimates aren't useful in improving perceived image quality. The estimate uncertainties are often sufficiently low to improve our perception. As the signal-to-noise ratio for the data increases, the estimate uncertainties decrease. Many images have sufficient SNR for useful estimates. While shadow regions have lower SNRs, their information content is lower (less light means less information). The increase in pixel estimate uncertainties is often imperceptible.
To get the best possible estimates exposure (shutter time and aperture) must be optimized.Gratuitous underexposure means lower SNR which in turn means inferior estimates for the missing pixels. Unimportant highlights should be intentionally overexposed. CCD sensors typically have higher artifact levels from overexposed highlights compared to CMOS sensors. Highlight overexposure is more risky.
These methods do not address all the disadvantages of using lower resolution sensors. This primary disadvantage is an inferior spatial sampling rate compared to higher resolution sensors.[1] This causes aliasing artifacts. These artifacts are typically minimized using optical filters which blur the image and reduces resolution. Sensors with higher sampling rates have more resolution so by now optical filters are unnecessary. A more subtle effect is the correlation of Gaussian noise between the real pixels used to estimate the missing pixels. For subjects with high levels of detail, high resolution sensor pixels are more informative in terms of SNR compared to estimated pixels that never existed in lower resolution sensors.
1/ link
There Are No Miracles
But that doesn't mean one shouldn't use methods that make the most out of their data.
I don't think Adobe is doing anything new or unique.
What Adobe has done is increase the feature set of Camera Raw. Resolution is not increased. A mathematical model is used to estimate the information content of pixels that never existed. There is no new information in the modeled image. Images created by these methods don't have more information content nor do they the same information content as a higher resolution sensor image.
Adobe, and others, are taking advantage of higher performance computers to render images with superior perceived image quality.
This does not the model's parameter estimates aren't useful in improving perceived image quality. The estimate uncertainties are often sufficiently low to improve our perception. As the signal-to-noise ratio for the data increases, the estimate uncertainties decrease. Many images have sufficient SNR for useful estimates. While shadow regions have lower SNRs, their information content is lower (less light means less information). The increase in pixel estimate uncertainties is often imperceptible.
To get the best possible estimates exposure (shutter time and aperture) must be optimized.Gratuitous underexposure means lower SNR which in turn means inferior estimates for the missing pixels. Unimportant highlights should be intentionally overexposed. CCD sensors typically have higher artifact levels from overexposed highlights compared to CMOS sensors. Highlight overexposure is more risky.
These methods do not address all the disadvantages of using lower resolution sensors. This primary disadvantage is an inferior spatial sampling rate compared to higher resolution sensors.[1] This causes aliasing artifacts. These artifacts are typically minimized using optical filters which blur the image and reduces resolution. Sensors with higher sampling rates have more resolution so by now optical filters are unnecessary. A more subtle effect is the correlation of Gaussian noise between the real pixels used to estimate the missing pixels. For subjects with high levels of detail, high resolution sensor pixels are more informative in terms of SNR compared to estimated pixels that never existed in lower resolution sensors.
1/ link
raid
Dad Photographer
WOW! A digital new revolution?
agentlossing
Well-known
I should try out some of my old Nikon D40 6MP files in it, could be interesting to play around with at least.
hap
Well-known
I have been ruminating on using Gigapixel AI for my Nikon 1 V1 files. Sounds like the same thing but built into PS....which I don't have.
Dralowid
Michael
...and I wonder what would it do to 35mm scans from an old Epson v500?
raid
Dad Photographer
What is the cost of such new software to get this feature?
PaulDalex
Dilettante artist
After seeing this thread I downloaded Bridge (BTW Adobe forces you to first install Adobe Creative Claud)
Then I tried to open in Bridge a photo using Camera Raw.
I got the message:
"Editing in Camera Raw is not enabled. Camera Raw editing requires that a qualifying product has been licensed"
THEREFORE CAMERA RAW IS NOT FREE AT ALL!
Then I tried to open in Bridge a photo using Camera Raw.
I got the message:
"Editing in Camera Raw is not enabled. Camera Raw editing requires that a qualifying product has been licensed"
THEREFORE CAMERA RAW IS NOT FREE AT ALL!
benlees
Well-known
Bridge is an organizing software, regardless. Elements is quite affordable- and often on sale. Camera Raw should be free if you purchase that.
PaulDalex
Dilettante artist
One can do the same enhance photo within PS but only for raw files, not jpegs.
So now the next challenge is to uninstall Bridge and Adobe Creative Cloud
Windows won't uninstall them
So now the next challenge is to uninstall Bridge and Adobe Creative Cloud
Windows won't uninstall them
PaulDalex
Dilettante artist
Hi benlees,
Elements is an 8 bit product. After such a long time since its introduction
I would not used it not even for free
Elements is an 8 bit product. After such a long time since its introduction
I would not used it not even for free
benlees
Well-known
Elements is 16 bits for RAW, 8 bits for jpeg. Honestly, and no disrespect, if you can tell the difference then your monitor is excellent and it may worth it for you to invest in "real" photoshop.
hap
Well-known
I've run some Epson files through Topaz Gigapixel AI and the results can be impressive.
Color differences are just from processing differences between the files.
Shawn
Shawn
Do you have an opinion regarding the effectiveness of Gigapixel AI to the new Super resolution in PS. Is Super Resolution functional in Elements?
hap
Well-known
Does Elements allow plug ins like LR or PS etc?
shawn
Veteran
I stopped upgrading Adobe once it went subscription based so I don't have the ability to compare. Gigapixel AI and Sharpen AI can be scary effective at times though.
For example this is from a DP3M. Left side is original image zoomed at 400%. Right side is a Gigapixel 2x upscale and Sharpen AI (in stabilize mode) viewed at 200%.
Shawn