Adobe Photoshop's New Super Resolution Feature is 'Jaw-Dropping' (petapixel.com) 146
Adobe just dropped its latest software updates via the Creative Cloud and among those updates is a new feature in Adobe Camera Raw (ACR) called "Super Resolution." You can mark this day down as a major shift in the photo industry, writes PetaPixel. From the report: I have seen a bit of reporting out there on this topic from the likes of PetaPixel and Fstoppers, but other than that the ramifications of this new feature in ACR have not been widely promoted from what I can see. The new Super Resolution feature in ACR essentially upsizes the image by a factor of four using machine learning, i.e. Artificial Intelligence (AI).
The PetaPixel article on this new feature quoted Eric Chan from Adobe:
Super Resolution builds on a technology Adobe launched two years ago called Enhance Details, which uses machine learning to interpolate RAW files with a high degree of fidelity, which resulted in images with crisp details and fewer artifacts. The term 'Super Resolution' refers to the process of improving the quality of a photo by boosting its apparent resolution," Chan explains. "Enlarging a photo often produces blurry details, but Super Resolution has an ace up its sleeve: an advanced machine learning model trained on millions of photos. Backed by this vast training set, Super Resolution can intelligently enlarge photos while maintaining clean edges and preserving important details."
What does this mean practically? Well, I immediately tested this out and was pretty shocked by the results. Though it might be hard to make out in the screenshot below, I took the surfing image shown below, which was captured a decade ago with a Nikon D700 -- a 12MP camera -- and ran the Super Resolution tool on it and the end result is a 48.2MP image that looks to be every bit as sharp (if not sharper) than the original image file. This means that I can now print that old 12MP image at significantly larger sizes than I ever could before. What this also means is that anyone with a lower resolution camera, i.e. the current crop of 24MP cameras, can now output huge image files for prints or any other usage that requires a higher resolution image file. In the three or four images I have run through this new feature in Photoshop I have found the results to be astoundingly good.
Super Resolution builds on a technology Adobe launched two years ago called Enhance Details, which uses machine learning to interpolate RAW files with a high degree of fidelity, which resulted in images with crisp details and fewer artifacts. The term 'Super Resolution' refers to the process of improving the quality of a photo by boosting its apparent resolution," Chan explains. "Enlarging a photo often produces blurry details, but Super Resolution has an ace up its sleeve: an advanced machine learning model trained on millions of photos. Backed by this vast training set, Super Resolution can intelligently enlarge photos while maintaining clean edges and preserving important details."
What does this mean practically? Well, I immediately tested this out and was pretty shocked by the results. Though it might be hard to make out in the screenshot below, I took the surfing image shown below, which was captured a decade ago with a Nikon D700 -- a 12MP camera -- and ran the Super Resolution tool on it and the end result is a 48.2MP image that looks to be every bit as sharp (if not sharper) than the original image file. This means that I can now print that old 12MP image at significantly larger sizes than I ever could before. What this also means is that anyone with a lower resolution camera, i.e. the current crop of 24MP cameras, can now output huge image files for prints or any other usage that requires a higher resolution image file. In the three or four images I have run through this new feature in Photoshop I have found the results to be astoundingly good.
Cool Advertising (Score:5, Interesting)
This has been part of the Topaz Labs ecosystem (a picture/video editor) for years. Adobe finally getting tech integration that's been common in the competition doesn't make for 'historic' days.
Re:Cool Advertising (Score:5, Insightful)
Who is Topaz Labs? The reason this is news is because nobody has ever heard of Adobe's competition.
Re: (Score:2)
Topaz isn't a competitor. It's a major suite of plugins for Adobe tools that are widely used in industry.
Re: (Score:2)
Re:Arent you a snowflake.... (Score:4, Informative)
They are a special purpose software company. The only way they compete with Adobe is in image resizing, image denoising, and video resizing. Needless to say I am thorroughly unimpressed with Adobe's new feature, and the amazement by Eric Chan in TFA just sort of plays to his ignorance. We've had the ability to do these kinds of resizes for at *least* 6 years now, ... Ironically Topaz's first foray into this enhancement started with a Photoshop plugin.
Re: (Score:2)
Why is it that no one is mentioning the actual big picture* here?
We are teaching software to imagine. To extrapolate what isn't there from what is, based on experience and educated guessing. This is exactly what intelligent creatures do with things like object permanence.
Sure this isn't actual "AI", but this sort of capability, once it has advanced enough, is a major building block of what proper AI needs to exist. It's fascinating, and slightly scary.
*pun not intended, but heartily embraced
Re: (Score:2)
I haven't looked, but some reviews put it way ahead of Topaz - which is surprising for a first release even if it is from someone as big as Adobe.
This is better (Score:4, Informative)
This has been part of the Topaz Labs ecosystem (a picture/video editor) for years.>
It has, but the Photoshop version seems better (fewer artifacts) and runs much faster (3 seconds vs. 1 minute 23 seconds) [fstoppers.com].
I also have Topaz Gigapixel, but the artifacts really put me off from using it much, and Photoshop seems to have done a much better job controlling it. Have not tried on my own images yet.
Re: (Score:2)
I know you couldn't be bothered to RTFA in your breathless rush to tell everyone how unimpressed you are, but it mentions Topaz:
Nobody is claiming Adobe invented AI image upscaling (and nobody used the word 'historic' except you). But Photoshop is THE industry standard, and them launching a best-in-class implementation of this feature WILL cause a major shift and mass adoption of this as stand
Re: Cool Advertising (Score:2)
Re: (Score:2)
Yeah.
Besides, I bet all those pictures are photoshopped anyway.
Re: (Score:2)
They're not - they're raw images of photoshopped objects. [xkcd.com]
Re:Cool Advertising (Score:5, Insightful)
And NVIDIA, for example, has DLSS. But this is the photo industry... they aren't the smartest bulbs, and they worship Adobe like Apple fans worship Mac, so it's entirely unsurprising that they see something commonplace as a holy grail.
Can you please tell me where I can download nvidia's DLSS tool to process images? I went to https://www.nvidia.com/en-us/g... [nvidia.com] to download it, but it seems it's not actually a tool that photographers can use as a standalone tool (with the annoying export/process/re-import steps), much less a plugin integrated into their standard editing workflow. It seems that's it's nothing more than a feature that software could implement (and they only seem to list games as the software that has implemented it). Furthermore, DLSS requires the latest series of nvidia graphics cards to run, as opposed to adobe's feature that, as far as I can determine, has the following minimal GPU requirements:
GPU with DirectX 12 support
2 GB of GPU memory
Which is really quite a lot more flexible. And even then, I'm not entirely clear if this is required, or if it can run 100% on the CPU (at a much slower rate) like pretty much all of photoshop's other GPU accelerated features.
So yeah, I think you intended to be quite clever with your snippy reply, but I think you failed miserably. You probably shouldn't be the one to be saying other people " aren't the smartest bulbs"
LOL...And just before I submitted this, I decided to look a little more into what DLSS is about.
https://www.nvidia.com/en-us/g... [nvidia.com]
DLSS 2.0 has two primary inputs into the AI network:
1. Low resolution, aliased images rendered by the game engine
2. Low resolution, motion vectors from the same images -- also generated by the game engine
Great...so tell me, which model of camera records a list of motion vectors for each object captured in the scene? Because those are the camera's that this would work on (I'll give you a hint....the number is very close to your IQ). Not the brightest bulb indeed. Oops, sorry, my mistake. The phrase you used was "smartest bulb".
Re: (Score:2)
DLSS could probably work with some additional effort on video. You'll just have to run it through a first pass to generate motion vectors. It does work significantly differently than the Super Resolution thing though, because it collects information across multiple frames to reconstruct the missing details, while Photoshop would have to completely invent stuff.
It'd be interesting to try it, but all the screenshots in the article seem to have been downscaled so it's hard to tell if it's really doing any good
Re: (Score:2)
I don't think so. Because in my brief reading about it (which I admit isn't super extensive) it appears to me that the way DLSS works is that it knows the exact geometric specification of each object and knows where it moved frame to frame, and then uses that to very precisely render different subpixels each frame, and then combine the results from frame to frame. It's almost like a 3d rendered form of interlaced video or progressively rendered images (but much better, obviously).
Yes, something similar can
Re: (Score:2)
Re: (Score:2)
www.topazlabs.com. Just because the GP used DLSS as an example which can't be applied simply to images, doesn't mean trained neural network based image resampling hasn't been available as a downloadable software package (or even Photoshop plugin) for a good 5+ years now.
There's nothing new here to those of us who take interest into the world outside what Adobe presents us. Hell if you were looking for quality resizing algorithms in the past you'd have to have been frigging stupid to simply accept what Adobe
Re: (Score:2)
Re: (Score:2)
Cool claim. The linked article's pictures so far don't back up that claim in the slightest, and I'm thorroughly unimpressed with the results that seem to be equal to what I've been getting for several years now.
Re: (Score:2)
DLSS is a bit different since it works in real time and requires per-game learning.
The real problem with DLSS though is despite repeated claims that they've made the results better each time I've tried it the results looked like shit and I turn it off.
Re: (Score:2)
DLSS no longer needs to be trained per game. It also doesn't look like shit (any more)
The current consensus from tech journalism, reviewers, and consumers on DLSS is that it can give dozens of FPS savings with nigh indistinguishable results versus a higher native resolution. It's at the point point where AMD's lack of DLSS flavor on their cards is a major competitive disadvantage at the moment from a value offer perspective.
And I'm getting an AMD card, so I'm not being a fanboy here. I'm just a AAA game dev
Re: (Score:2)
For someone who just wants to be able to blow up holiday photos for printing, does Topaz' software hold up? I don't do subscription software, so Adobe is simply a non-starter.
Or is there something slower and perhaps 80% of the quality available (even commercially) for GIMP?
JAW DROPPING!!!!! (Score:5, Insightful)
But not jaw dropping enough for me to rent software.
Re: (Score:2)
I had a look for an open source version of this yesterday. There isn't really anything.
There are some specialist ones for anime that look cool, but something I can drop an arbitrary photo onto and get back a 2x resolution version... Nothing as far as I can tell.
Linus Tech Tips has a lengthy video (Score:2)
The big thing was how everything integrates. Adobe has a product for everything. You can find cheaper alternatives that you can own, but then you're constantly fussing about to get them all working together. It doesn't help that his employees are all used to Adobe.
The lost time wasn't worth it. This isn't a surprise. Adobe's a huge company. They'd do studies to show how much money people make and save usin
Re: (Score:2)
If you want to actually buy software that does this better than photoshop does then Topaz Labs is more than happy to sell you their image resizer which has done this for many years already.
This is going places (Score:2)
1. "Enhance" grainy security footage with generated pixels
2. Match generated pixels against facial recognition data base to produce generated perp
3. Kill perp in the course of serving a SWAT style no-knock warrant
4. CASE CLOSED (profit?)
Re:This is going places (Score:5, Interesting)
I have always wanted to see a scene in a cop show involving grainy, low res security video:
Detective: "Can you enhance that so we can run facial rec on it?"
Tech: "Sure thing, boss. Who do you want it to look like?"
Re: (Score:2)
Enhance 224 to 176.
Enhance, stop.
Move in, stop.
Pull out, track right, stop.
Center in, pull back. Stop. Track 45 right. Stop.
Center and stop. Enhance 34 to 36.
Pan right and pull back. Stop.
Enhance 34 to 46. Pull back. Wait a minute, go right, stop.
Enhance 57 to 19. Track 45 left. Stop.
Enhance 15 to 23.
Give me a hard copy right there.
Enhance you say? (Score:3)
Did Adobe take a cue from Red Dwarf [youtube.com]?
Re: (Score:2)
That was a rather blatant riff on Blade Runner: https://youtu.be/hHwjceFcF2Q [youtu.be]
Re: (Score:2)
The "uncrop" feature sure would be handy.
Re: (Score:2)
Adobe Photoshop has that. It extrapolates based on the scene contents and algorithmically expands. Added a decade ago in CS6 and it is much more sophisticated these days.
Yay (Score:2)
Finally crimes can be solved by zooming in on a reflection on someoneâ(TM)s eyeball.
Better yet, actually better than existing solution (Score:2)
I think a better link to see how well the new super-resolution works, is this Fstoppers [fstoppers.com] story comparing the Photoshop update to an existing solution, Topaz Gigapixel - the Photoshop results were faster, with significantly fewer artifacts.
But did you though? (Score:4, Insightful)
the end result is a 48.2MP image that looks to be every bit as sharp (if not sharper) than the original image file. This means that I can now print that old 12MP image at significantly larger sizes than I ever could before.
But did you actually print the image at a significantly larger size and compare the results?
Re: (Score:2)
Why? Do you think somehow a higher resolution picture that looks better on a screen will somehow have the opposite effect on the print?
If I have two hot plates on the oven and one is glowing red and the other is black and hasn't been turned on, does I really need to touch both of them and measure the severity of the burn to know which one of them is hotter?
Re: (Score:2)
Why? Do you think somehow a higher resolution picture that looks better on a screen will somehow have the opposite effect on the print?
I question whether a 48 megapixel image is going to appear noticeably better than a 12 megapixel image if they're being viewed on a computer monitor. Even a 4K UHD screen can only display around 8 megapixels. You either need to downscale the image to fit on your screen, or scroll around looking at small portions of the image. In which case, what is the point?
Re: (Score:2)
Why? Do you think somehow a higher resolution picture that looks better on a screen will somehow have the opposite effect on the print?
I question whether a 48 megapixel image is going to appear noticeably better than a 12 megapixel image if they're being viewed on a computer monitor. Even a 4K UHD screen can only display around 8 megapixels. You either need to downscale the image to fit on your screen, or scroll around looking at small portions of the image. In which case, what is the point?
Cropping is the point. You can adjust your image composition to a much higher degree.
Also, printouts at a big size also need lotsa pixels.
For what? (Score:5, Interesting)
Professional photographers generally have a camera sensor with a resolution that's already excessive for whatever they're doing. The only exceptions I can think of, where higher resolutions would be useful, would be scientific or historical archiving purposes...for which you definitely don't want AI taking educated guesses at what's supposed to be there.
Re: (Score:2)
How about non-professional photographers? :-)
Re: (Score:3)
Most don't know what resolution is and can't even hold the phone the right way up :-P
Re: (Score:2)
Still useful (Score:2)
Even if you have a 50MP camera, there are times when it's useful to have a good super scaler - if you end up wanting to heavily crop something for example.
For stock photographers, this could bring new life to older images.
Re: (Score:2)
Professional photographers generally have a camera sensor with a resolution that's already excessive for whatever they're doing.
I have several 32"x48" fine-art photography prints hung up in my house, printed on acrylic dibond. They look fine from a distance of a few feet, but they're the sort of pictures that invite close viewing, where I start to see artifacts. I'd be delighted for crisper edges through machine learning.
I don't know how to explain it. I believe the source cameras are either 36 megapixels (7k*4k which should have yielded 125 pixels per inch at that size) or 50 megapixels (9k*6k which should have yielded 200 pixels p
Very Common in Science (Score:2)
The only exceptions I can think of, where higher resolutions would be useful, would be scientific or historical archiving purposes...for which you definitely don't want AI taking educated guesses at what's supposed to be there.
What are you talking about? There is nothing wrong with an educated guess in science provided that you can quantify the uncertainty on the process used to reach that educated guess.
We do this all the time in science because what we measure is reality as distorted by our detectors and instruments. We then use a variety of statistical methods - increasingly including machine learning algorithms - to make an estimate (i.e. educated guess) of what the real distribution was before our detectors mangled it. T
Re: (Score:2)
Yes, the analysis part is the difference, you wouldn't want Adobe's black box doing who-knows-what with your source data.
Re: (Score:2)
would be scientific or historical archiving purposes
Oh I feel sorry for you. Your family discovered digital cameras only after fancy high res sensors came out. That must suck
In the meantime there are people all over the world who aren't scientists or museum curators who have low resolution images on their PCs which they would like to blow up without nasty pixelation as a result.
Re: (Score:2)
What is USUALLY there (Score:2)
AI is trained to interpolate based on what is normally there. We ought to remember to not trust such technique(s) in unusual images, such as those of other planets.
They should also be scrupulously analyzed, if the results are used in court-proceedings as evidence...
Re: (Score:2)
This applies equally true to any digital capture even straight out of the hardware. There's a huge amount of interpretation of lighting and range-finding data that goes into recording the eventual .jpg file. And the old analog cameras? Same problem, but it was more the human being behind the shutter doing the biasing with the F-stop and shutter speed. The raw film out of the camera is already the result of tons of intervention by the observer. So for court cases, be careful. The problem of the OJ Simpson ma [wikipedia.org]
Re: (Score:2)
But those are fairly straightforward techniques, aren't they, not "artificial intelligence", the workings of which aren't always understood even by the very programmers creating it.
So it make it up, then? (Score:3)
I've seen some of these things in action and they blow you away for a while. Once the novelty wears off you realise that the "machine learning" isn't perfect and is, indeed, making it up. Eventually, the draw of fidelity proves stronger.
Also, PetaPixel are the same sort of retards who used to write for Hi-Fi magazines. Their opinions have no value to the reader, only to whoever paid for them.
Cell Phones Doing this for Years (Score:2)
Re: Cell Phones Doing this for Years (Score:2)
iPhones dont automatically upscale
Re: (Score:2)
Test it properly (Score:2)
Please someone scale up the original Duke Nukem face. Then feed it back a bunch of times. Then add a link so I get to see it in full res :-)
Finally! (Score:2)
Now that actually printed ads are as good as dead, they come with this feature.
For all of us that want to create large prints (Score:2)
I'm willing to bet that 99% of photographers will not produce a print larger than 4' (on an edge), seeing as how many large format printers do about 180PPI a 50MP camera will do well enough. For those of us who want make even bigger pictures, sure up-scaling could make things better. But for me it's not earth shattering and I could already upscale with topaz labs.
Cool! (Score:2)
Now add this to browsers so that we only need to send lightweight 320x240 JPEGs and pass them through this algorithm recursively to get HiDPI images displayed on websites.
How much larger do you need? (Score:2)
12MP contains more pixels than necessary to print a 1200dpi poster size (24 inches by 36 inches) print.
48MP is enough to print something the size of the side of a billboard.
That's besides the fact that all these extra pixels are just machine interpreted pixel doubling, not actual data. That's like saying your pixel doubling 4K TV is providing more data when it's not.
The T-shirt on the surfer in the image (Score:2)
What about super duper resolution ? (Score:2)
Put it in a loop and before you know it you have 12 Gigapixel image. Image of what I don't know, but it will be 12 GP for sure...
I think 12MP is "low resolution" these days (Score:3)
I tried this on several really low resolution photos from early 2000s. (Crappy cameras, or scanned film based photos). And it only adds "noise" to the photo where there was more of a blur before. It just does not seem to work under those conditions.
Yes, it might make your 12MP into a 48MP. However you can already print the 12MP in very large formats.
Re: The illusion of quality, the sickness of quant (Score:3)
The reality is your camera is introducing artifacts into the picture. If the AI is removing those artifacts, then nothing of value is lost and the photos are more real than otherwise.
Re: (Score:2)
Re: The illusion of quality, the sickness of quan (Score:2)
Increasing resolution is trivial to do without artifacts. You can't assume this process is introducing artifacts. It probably is, but it's not given.
Re:The illusion of quality, the sickness of quanti (Score:5, Insightful)
What nonsense. Photography is just a technology, the point of photography is whatever the photographer wants it to be. That can be art, preserving reality, or anything. And who cares about "fake pixels" if the result is indistinguishable?
Re: (Score:3)
And let's not even talk about preserving reality in an article about photoshop!
You need to think professionally (Score:2)
Tech like this being widely available means you can use cheaper cameras. That in turn will open up more amateurs to go pro (which depending on your perspective might not be a good thing, since it could flood the market).
It's also another thing that used to be done painstakingly by hand (cleaning up a low res image to make it suitable for high res printing) that can be done by AI now.
Re: (Score:2)
"the point of photography is whatever the photographer wants it to be"
But that's not really true. It is a combination of what the photographer wants and what the viewer wants and expects. Most viewers look at photography differently that other visual arts and this creates context-sensitive expectations and ethics. For example, photography is certainly not `whatever the photographer wants` in a news context and photographers often get fired thinking that.
Re: The illusion of quality, the sickness of quant (Score:2)
While I agree with you and think what you do is awesome, this is a bit different because its not just changing the colour to make it look more like this or that reality, its actually fabricating reality. Some of the details rendered will not have actually been there when the shot was taken.
No doubt the pictures will look great but they arent entirely a capture of reality. They are partly fake. I think thats what the original poster was fumbling trying to say.
Re:The illusion of quality, the sickness of quanti (Score:5, Informative)
The point of photography is to record reality.
That's just one of many possible motivations....
This idea that you should vomit out fake pixels to fool people and "fill in" what wasn't there is disgusting.
Ever heard of demosaicing? If not you're in for a bit of a shock.
Re:The illusion of quality, the sickness of quanti (Score:5, Insightful)
The point of photography is to record reality
Tell that to a wedding photographer. Photography hasn't been about recording reality for well over a century. Photography is also used as an artistic story to make an impression or tell a story.
This technology should not be used on security camera footage for a court case. That's not its intended purpose - even if Adobe literally named the menu entry "Enhance."
Re: (Score:2)
Photography is also used as an artistic story to make an impression or tell a story.
Tell that to a street photographer, a documentary photographer, a photojournalist, etc. etc. etc. Oh I suppose this tells a story, but without integrity for these types of photographers. Sorry, but you are even more narrow minded than those that you make accusations about.
Re: (Score:2)
You missed a key word of "also." Nobody is holding a gun to your head telling you to click that button.
Re: (Score:2)
Given that your brain works in real time much the same way, according to your theory everything you see is fraud.
Re: (Score:2)
Even more so. The bandwidth of the optic nerve is shockingly low.
Re: (Score:2)
Fraud, lol. I'm interested in an open source version of this so that I can improve photos taken with my phone to the point where they look really nice on a 4k display, with a bit of cropping.
I don't care if the image isn't a perfect record of real life. The phone already processed the hell out of it. I just want it to look nice.
Re: (Score:2)
Photographers were touching up, enhancing and otherwise manipulating photographs in the darkroom for decades before computers even existed [metmuseum.org]. Sometimes it was with the purpose of lying [wikipedia.org], but other times it was for pure artistic expression [wikipedia.org], or to meet the needs of advertising [wikimedia.org].
Re: (Score:2)
Re: (Score:2)
Re: The illusion of quality, the sickness of quant (Score:2)
hehe I agree with you but the same argument can be made for interpolated frames.
Re: (Score:2)
The point of photography is to record reality. You are capturing existence as it was for a moment, through a visual record.
Just curious...are you against zoom lenses, too?
Re: (Score:2)
The point of photography is to record reality.
I have a book published in the 50s showing a photograph, black and white of course, of a scuba diver emerging from a street puddle. Why do I say this? Well I want to point out that when I say you're incredibly ignorant I don't just mean ignorant of modern photography, but I mean ignorant of photography as the art form has always existed. An ignorance which transcends time itself. That's really quite impressive.
Now go down to an op-shop and pickup a book on how to use a darkroom for $1 if you're ever actuall
Re: (Score:2)
> The point of photography is to record reality.
1. You DO realize that every medium is an approximation, right?
2. Do you also dismiss some of the amazing Black and White photographs [webneel.com] as well?
> This idea that you should vomit out fake pixels to fool people and "fill in" what wasn't there is disgusting.
You mean the same way chemical residue on film does?
Or ink on paper does?
Why do you assume that fancy 10-color ink being not only matches but perfectly replicates the actual frequency of the original photo
Re: The illusion of quality, the sickness of quant (Score:2)
I agree with you and am not backing up the poster youre replying to, but I think a case can be made that, at least in some contexts, there is a difference between attempting to capture or render an image as closely to what the eye sees as possible and deciding to intentionally fabricate content.
In many cases the upscaled images will look virtually identical to one which had been taken at that resolution in the first place, but in many it wont and I think sometimes that will matter. Sometimes it will be art
You are correct! (Score:2)
Many people above are arguing with you about the true purpose of photography which is an opinion, but you are completely correct that super resolution is inserting fake pixels.
The summary states, "Super Resolution can intelligently enlarge photos while maintaining clean edges and preserving important details." But this is not true! The software is not preserving details, it is inserting details that it thinks are most likely based on the realism that it has learned from millions of other images.
People b
Re: You are correct! (Score:2)
Exactly. Some here are overblowing the ramifications of that, but others are hand waving it away. Its important to be aware of.
Re: (Score:2)
Your camera downsamples reality (nevermind introduces artifacts of varying types). So anything it "reports" isn't exactly reality either.
Re: (Score:2)
Re: (Score:2)
Don't kid yourself -- film straight out of the camera has a whole bunch of issues just like digital.
Re: (Score:2)
First, I don't think 'typical consumers' use Photoshop. That aside, this sort of thing is important for enlargements. Say you wanted to 'zoom in' on just the surfers face, and blow that up to the size of the original picture. Now higher resolution becomes very important.
Re: (Score:2)
I can see plenty of potential usages. If you are trying to mix and match images at different resolution at could look funky and so inventing the "middle pixels" could help removing the upscaling artifacts.
Re: (Score:2)
If you're printing for billboards and display banners, you also need really high resolution, particularly if the banner will be seen from far away but people will be walking up close to it, as might happen in sidewalk displays.
Re: (Score:2)
Re: Only works with Camera RAW files (Score:2)
Yes, iPhones.
Re: Only works with Camera RAW files (Score:2)
Re: (Score:2)
Nonsense. You can use an appropriately trained upscaling technique on anything you want. I don't even think there would be any significant benefit to using the raw data since de-Bayering is a lossless and reversible operation. What you lose converting to a JPEG is bit depth, but that's not going to have much effect on upscaling.
Re: (Score:2)
Yeah, exactly what I was thinking. Zoom in and enhance. [Repeat 80 times]