Enhance, Rotate, Zoom in.
April 12, 2011 5:07 PM Subscribe
An amazing bit of photoshoppery transforms an extreme oblique view of a poster into a straight on view.
I just did the same thing with his source image (that's listed at 100%) and got this. Either he's working with a MUCH higher resolution photo, this is a lie, or it's a joke I just don't get. Any way you slice it, though, it's a shitty photoshop tutorial.
posted by CharlesV42 at 5:25 PM on April 12, 2011 [1 favorite]
posted by CharlesV42 at 5:25 PM on April 12, 2011 [1 favorite]
This one says 25%. Someone who cared to could likely use known sizes for Mac/photoshop GUI elements and extrapolate the real size of the original image.
Your result seems remarkably close to his, I think. Start with an 8 or 10 megapixel image instead of a 400x551 regular old pixel image, and you'd be golden.
posted by jsturgill at 5:31 PM on April 12, 2011
Your result seems remarkably close to his, I think. Start with an 8 or 10 megapixel image instead of a 400x551 regular old pixel image, and you'd be golden.
posted by jsturgill at 5:31 PM on April 12, 2011
He also did the wrong way. You should correct lens distortion first, and then take care of the perspective. This is why he still has curved edges to his poster.
posted by Pyry at 5:35 PM on April 12, 2011 [10 favorites]
posted by Pyry at 5:35 PM on April 12, 2011 [10 favorites]
I don't get it, but then again, I suck at Photoshop.
posted by benito.strauss at 5:35 PM on April 12, 2011 [1 favorite]
posted by benito.strauss at 5:35 PM on April 12, 2011 [1 favorite]
you'd be surprised at really how not-very-amazing a human doing this with photoshop is. distortion and perspective manipulation is covered in most intro to photoshop courses.
however, i continue to be amazed at what our technologies can do, so kudos for finding delight in this blade-runner-like feature of a popular software package
posted by Señor Pantalones at 5:39 PM on April 12, 2011 [7 favorites]
however, i continue to be amazed at what our technologies can do, so kudos for finding delight in this blade-runner-like feature of a popular software package
posted by Señor Pantalones at 5:39 PM on April 12, 2011 [7 favorites]
So, like, the side-on view of a picture got turned around a bit so it was, like, full on?
I am trying to be impressed here. Perhaps this is something one would have to be a geek to appreciate.
posted by Decani at 5:40 PM on April 12, 2011
I am trying to be impressed here. Perhaps this is something one would have to be a geek to appreciate.
posted by Decani at 5:40 PM on April 12, 2011
Any way you slice it, though, it's a shitty photoshop tutorial.
This reaction seems really strange to me. First of all, you got pretty amazing results from following his supposedly shitty tutorial, and second, it seems to me rather obvious that the difference is accounted for in the relative resolutions of your starting images. So I have no idea what you're (apparently) angry about.
posted by neuromodulator at 5:40 PM on April 12, 2011 [6 favorites]
This reaction seems really strange to me. First of all, you got pretty amazing results from following his supposedly shitty tutorial, and second, it seems to me rather obvious that the difference is accounted for in the relative resolutions of your starting images. So I have no idea what you're (apparently) angry about.
posted by neuromodulator at 5:40 PM on April 12, 2011 [6 favorites]
Up next: convert a color image to greyscale in 16 easy steps!
posted by ShutterBun at 5:40 PM on April 12, 2011 [6 favorites]
posted by ShutterBun at 5:40 PM on April 12, 2011 [6 favorites]
Perhaps this is something one would have to be a geek to appreciate.
Perhaps this is one of the times when you could not post, instead.
posted by neuromodulator at 5:41 PM on April 12, 2011 [1 favorite]
Perhaps this is one of the times when you could not post, instead.
posted by neuromodulator at 5:41 PM on April 12, 2011 [1 favorite]
Perhaps this is one of the times when you could not post, instead.
posted by neuromodulator at 1:41 AM on April 13
It seems a little late for that, I fear.
posted by Decani at 5:42 PM on April 12, 2011
posted by neuromodulator at 1:41 AM on April 13
It seems a little late for that, I fear.
posted by Decani at 5:42 PM on April 12, 2011
You should correct lens distortion first, and then take care of the perspective.
I love learning something new. I have been working on improving my photoshop skills, so I found this post interesting. Hopefully we can salvage this thread with more insight, and less "that was the dumb!"
posted by archivist at 5:44 PM on April 12, 2011 [2 favorites]
I love learning something new. I have been working on improving my photoshop skills, so I found this post interesting. Hopefully we can salvage this thread with more insight, and less "that was the dumb!"
posted by archivist at 5:44 PM on April 12, 2011 [2 favorites]
So we're one step closer to Photoshop CSI?
posted by maudlin at 5:45 PM on April 12, 2011 [4 favorites]
posted by maudlin at 5:45 PM on April 12, 2011 [4 favorites]
I've actually done this for an image used a professional publication. There was a painting we had permission to reproduce, but the person who took the photos was an amateur and didn't get a shot anywhere near straight-on, although not as extreme as the one in the link.
I would have just had someone re-shoot it (or do it myself), but the painting was in another country halfway around the world and the photographer wasn't there anymore. While it was legal for us to reproduce it, it was just easier for me to fix the distortion in Photoshop rather than deal with the language barrier and find someone local to shoot it.
Luckily, the digital camera used was high-enough resolution for me to be able to do this, plus the lighting and focus was also even enough that you'd never know how skewed the original was.
posted by D.C. at 5:49 PM on April 12, 2011 [1 favorite]
I would have just had someone re-shoot it (or do it myself), but the painting was in another country halfway around the world and the photographer wasn't there anymore. While it was legal for us to reproduce it, it was just easier for me to fix the distortion in Photoshop rather than deal with the language barrier and find someone local to shoot it.
Luckily, the digital camera used was high-enough resolution for me to be able to do this, plus the lighting and focus was also even enough that you'd never know how skewed the original was.
posted by D.C. at 5:49 PM on April 12, 2011 [1 favorite]
It seems a little late for that, I fear.
I'll photoshop it out.
posted by neuromodulator at 5:49 PM on April 12, 2011
I'll photoshop it out.
posted by neuromodulator at 5:49 PM on April 12, 2011
This post provides nothing for me to be outraged about.
posted by killdevil at 5:50 PM on April 12, 2011 [18 favorites]
posted by killdevil at 5:50 PM on April 12, 2011 [18 favorites]
Just to explain why there's a bunch of scorn here to those readers who aren't familiar with Photoshop, this isn't any kind of new feature or technological advance. Photoshop has been doing this for years with the same tools. Heck, it's basically just a digital version of the perspective manipulations photographers have been doing for ages with large format view cameras.
Nothing wrong with all that and it's a perfectly nice demo of what we can do, but that's the reason we're not so amazed.
posted by zachlipton at 5:51 PM on April 12, 2011 [4 favorites]
Nothing wrong with all that and it's a perfectly nice demo of what we can do, but that's the reason we're not so amazed.
posted by zachlipton at 5:51 PM on April 12, 2011 [4 favorites]
I dunno about you guys, but I was pretty impressed. Strike that, I was floored. Not at the guy's skills -- he didn't do anything complicated in the least -- but just that it was possible at all. I never in a million years would have guessed that the original ultra-squashed image had enough visual information in it to look as natural as it did when stretched back out, and I would've sworn it was a fake if not for seeing CharlesV42's attempt.
posted by rifflesby at 5:52 PM on April 12, 2011 [13 favorites]
posted by rifflesby at 5:52 PM on April 12, 2011 [13 favorites]
I am constantly amazed at how much digital information is hiding in plain sight sometimes.
I don't think I could recover this from this with my film/paper processing skills when they were at their best. But with today's technology, I can accomplish the task with just a couple of clicks. I was impressed with his work; I wouldn't have even tried, assuming there wasn't enough information stored in the original to produce the final product.
posted by nomisxid at 5:53 PM on April 12, 2011 [4 favorites]
I don't think I could recover this from this with my film/paper processing skills when they were at their best. But with today's technology, I can accomplish the task with just a couple of clicks. I was impressed with his work; I wouldn't have even tried, assuming there wasn't enough information stored in the original to produce the final product.
posted by nomisxid at 5:53 PM on April 12, 2011 [4 favorites]
Fair enough. To reciprocate, as to why I found it impressive: I knew Photoshop has had all these tools for some time, but I would have guessed that there wasn't enough information in the starting image to reproduce the original so faithfully. I actually found it surprising.
Or what nomisxid said.
posted by neuromodulator at 5:54 PM on April 12, 2011
Or what nomisxid said.
posted by neuromodulator at 5:54 PM on April 12, 2011
Also, I think you can accomplish the same result by straightening the image, grabbing the crop tool, turning on the perspective option, dragging the crop handles to the edges of the poster, and hitting return.
♥ the perspective crop tool
posted by zachlipton at 5:55 PM on April 12, 2011 [2 favorites]
♥ the perspective crop tool
posted by zachlipton at 5:55 PM on April 12, 2011 [2 favorites]
I'll note that a big part of why this works is because the original is both absolutely flat, and has uniform areas of color. More organic and detailed images with significant local variation in the colors won't work out so well, neither would something like an oil painting with significant texture. When you get to something like a brick wall it totally changes.
posted by meinvt at 6:05 PM on April 12, 2011 [2 favorites]
posted by meinvt at 6:05 PM on April 12, 2011 [2 favorites]
Zachlipton has hit the nail on the head. It's a perfectly cromulent tutorial, but it's very, very basic stuff, and I kept getting the feeling that the joke was on me. Kinda felt like I was reading a Robert Benchley piece, where the delivery is so straight-faced that it takes a while to realize you've been had.
In this case it was just a nice little Photoshop 101 tutorial, so I was confused when the "amazing" didn't happen for me.
posted by ShutterBun at 6:13 PM on April 12, 2011 [1 favorite]
In this case it was just a nice little Photoshop 101 tutorial, so I was confused when the "amazing" didn't happen for me.
posted by ShutterBun at 6:13 PM on April 12, 2011 [1 favorite]
He also did the wrong way. You should correct lens distortion first, and then take care of the perspective. This is why he still has curved edges to his poster.
This.
Also, once you've done that, you'll mess up the aspect ratio for sure if you just eyeball the distort tool. The proper way is to figure out the original aspect ratio of the piece, set guidelines proportional to whichever side you've rotated to be vertical, then use the polygonal lasso or the pen tool to define the corners of the original poster. Drag those corners to the corners of the guidelined box, check that everything lines up, then crop.
I do this all the time preparing photographed paintings for websites.
Also, he sucks at the Curves tool. When you've got regional shadow like that, the answer is not that you should try to blow it out with a global adjustment. A better approach would be to apply the curve as a smart filter so that the adjustment behaves like a layer. Then all you have to do is layer mask the curves 'layer,' open the mask's alpha channel, and 'paint' the curves adjustment onto the affected areas with a large, soft, low opacity brush. That way you retain editability on your curves parameters, and you aren't just telling it "for every pixel approximately this dark, I want it this dark instead.
posted by fifthrider at 6:16 PM on April 12, 2011 [11 favorites]
This.
Also, once you've done that, you'll mess up the aspect ratio for sure if you just eyeball the distort tool. The proper way is to figure out the original aspect ratio of the piece, set guidelines proportional to whichever side you've rotated to be vertical, then use the polygonal lasso or the pen tool to define the corners of the original poster. Drag those corners to the corners of the guidelined box, check that everything lines up, then crop.
I do this all the time preparing photographed paintings for websites.
Also, he sucks at the Curves tool. When you've got regional shadow like that, the answer is not that you should try to blow it out with a global adjustment. A better approach would be to apply the curve as a smart filter so that the adjustment behaves like a layer. Then all you have to do is layer mask the curves 'layer,' open the mask's alpha channel, and 'paint' the curves adjustment onto the affected areas with a large, soft, low opacity brush. That way you retain editability on your curves parameters, and you aren't just telling it "for every pixel approximately this dark, I want it this dark instead.
posted by fifthrider at 6:16 PM on April 12, 2011 [11 favorites]
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.
posted by porpoise at 6:18 PM on April 12, 2011 [7 favorites]
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.
posted by porpoise at 6:18 PM on April 12, 2011 [7 favorites]
If you guys think this is amazing, you are going to poop your pants when you see the things Adobe introduced over the last 5 years.
posted by pokermonk at 6:25 PM on April 12, 2011
posted by pokermonk at 6:25 PM on April 12, 2011
While I appreciate a Robert Benchley reference as much as anyone, I'm going to stick by a "bullshit" assessment until someone here can actually replicate it using the given image, unless that's a massively downsampled version of what he started with there just aren't enough pixels there.
posted by DU at 6:27 PM on April 12, 2011
posted by DU at 6:27 PM on April 12, 2011
zachlipton writes "Just to explain why there's a bunch of scorn here to those readers who aren't familiar with Photoshop, this isn't any kind of new feature or technological advance. Photoshop has been doing this for years with the same tools. Heck, it's basically just a digital version of the perspective manipulations photographers have been doing for ages with large format view cameras. "
I hear what you are saying. I still find a child like wonder that it is even possible. I'm in awe of people running four minute miles and guys who can plane a 6X12 foot table top flat with nothing but a jointer (something tradesmen have been doing for at least a few centuries). Maybe my wonder bar is set too low.
pokermonk writes "If you guys think this is amazing, you are going to poop your pants when you see the things Adobe introduced over the last 5 years."
Please give us an example of the most amazing thing you can think of that can be done with photoshop. Unless it's being able to get a fingerprint of the reflection of a hand in a licence plate holder acron nut; I'd perfer to remain jaded about that piece of hollywood.
posted by Mitheral at 6:29 PM on April 12, 2011 [3 favorites]
I hear what you are saying. I still find a child like wonder that it is even possible. I'm in awe of people running four minute miles and guys who can plane a 6X12 foot table top flat with nothing but a jointer (something tradesmen have been doing for at least a few centuries). Maybe my wonder bar is set too low.
pokermonk writes "If you guys think this is amazing, you are going to poop your pants when you see the things Adobe introduced over the last 5 years."
Please give us an example of the most amazing thing you can think of that can be done with photoshop. Unless it's being able to get a fingerprint of the reflection of a hand in a licence plate holder acron nut; I'd perfer to remain jaded about that piece of hollywood.
posted by Mitheral at 6:29 PM on April 12, 2011 [3 favorites]
While I appreciate a Robert Benchley reference as much as anyone, I'm going to stick by a "bullshit" assessment until someone here can actually replicate it using the given image, unless that's a massively downsampled version of what he started with there just aren't enough pixels there.
As already noted, the source image is much larger than 1-to-1 pixels, as you can tell form the 25% zoom in the title of the actual manipulation screenshots.
posted by cortex at 6:30 PM on April 12, 2011
As already noted, the source image is much larger than 1-to-1 pixels, as you can tell form the 25% zoom in the title of the actual manipulation screenshots.
posted by cortex at 6:30 PM on April 12, 2011
I refuse to look at this unless it is translated into rap form.
posted by RobotVoodooPower at 6:31 PM on April 12, 2011
posted by RobotVoodooPower at 6:31 PM on April 12, 2011
I use photoshop. photoshop is an app of mine. This, sir, is no photoshop.
I'm calling shenanigans. The original photo is blurry, and yet, when he changes the scale, it's magically crisp. No way. Total bullshit.
posted by crunchland at 6:35 PM on April 12, 2011 [1 favorite]
I'm calling shenanigans. The original photo is blurry, and yet, when he changes the scale, it's magically crisp. No way. Total bullshit.
posted by crunchland at 6:35 PM on April 12, 2011 [1 favorite]
you can tell form the 25% zoom in the title of the actual manipulation screenshots.
But the image right above that says 100%. I guess that screenshot itself could be shrunk. It'd be nice if the guy posted the original. That's the point of a tutorial, right? To be able to try it yourself?
posted by DU at 6:38 PM on April 12, 2011
But the image right above that says 100%. I guess that screenshot itself could be shrunk. It'd be nice if the guy posted the original. That's the point of a tutorial, right? To be able to try it yourself?
posted by DU at 6:38 PM on April 12, 2011
Mitheral, have you seen context-aware fill? it's not as great in a fundamental tool way as these are, but it gives as good demo as the stamp tool did 15 years ago.
posted by bonaldi at 6:38 PM on April 12, 2011
posted by bonaldi at 6:38 PM on April 12, 2011
I still think content aware fill is neat. Brief example. Longer example.
posted by maudlin at 6:39 PM on April 12, 2011
posted by maudlin at 6:39 PM on April 12, 2011
Anyone with a camera and a nice flat painting can go try this at home right now. There's nothing magical about that Wonder Woman painting that makes that particular source image necessary.
posted by cortex at 6:43 PM on April 12, 2011
posted by cortex at 6:43 PM on April 12, 2011
THIS LOOKS SHOPPED
I CAN TELL FROM SOME OF THE PIXELS AND
FROM SEEING QUITE A FEW SHOPS IN MY TIME
posted by obiwanwasabi at 6:46 PM on April 12, 2011 [10 favorites]
I CAN TELL FROM SOME OF THE PIXELS AND
FROM SEEING QUITE A FEW SHOPS IN MY TIME
posted by obiwanwasabi at 6:46 PM on April 12, 2011 [10 favorites]
People. You're looking at web-friendly JPGs. The files he worked on were much higher res. This isn't difficult.
posted by John Kenneth Fisher at 6:52 PM on April 12, 2011 [4 favorites]
posted by John Kenneth Fisher at 6:52 PM on April 12, 2011 [4 favorites]
Skeptics and believers -- check out Hany Farid's work on image analysis and computer vision. Two neat examples: reconstructing Egyptian tomb artwork (on flat and curved walls) and distortion-free renderings of Orozco murals from photos taken at sub-optimal angles.
Pretty rad stuff.
posted by Westringia F. at 6:53 PM on April 12, 2011
Pretty rad stuff.
posted by Westringia F. at 6:53 PM on April 12, 2011
COMPUTER, ENHANCE
posted by regicide is good for you at 6:53 PM on April 12, 2011
posted by regicide is good for you at 6:53 PM on April 12, 2011
WHY AM I NOT BEING DAZZLED PROPERLY
posted by DoctorFedora at 7:03 PM on April 12, 2011
posted by DoctorFedora at 7:03 PM on April 12, 2011
(As penance for my hamfisted attempt to make a joke that was already made more skilfully, I offer, in the same vein: Crime Cops)
posted by regicide is good for you at 7:04 PM on April 12, 2011 [1 favorite]
posted by regicide is good for you at 7:04 PM on April 12, 2011 [1 favorite]
Anyone with a camera and a nice flat painting can go try this at home right now. There's nothing magical about that Wonder Woman painting that makes that particular source image necessary. --- Well, except that the picture on my wall that I take an oblique image of probably isn't up on flickr.
posted by crunchland at 7:12 PM on April 12, 2011
posted by crunchland at 7:12 PM on April 12, 2011
Evidently we've inspired an update:
"Update: People on the web have been posting confused unhappiness with a detail of the short demo that I posted here. Gotta address it. Please understand that I did all the photoshop work on the original picture. 8 megapixels for the original shot. NOT on the compressed, reduced JPG images that I posted here. I mean, seriously folks, I show screenshots taken from my (big) monitor. It should be clear that the image I'm working with isn't 400 pixels wide, eh? OK, on with the show."
posted by ShutterBun at 7:22 PM on April 12, 2011 [1 favorite]
"Update: People on the web have been posting confused unhappiness with a detail of the short demo that I posted here. Gotta address it. Please understand that I did all the photoshop work on the original picture. 8 megapixels for the original shot. NOT on the compressed, reduced JPG images that I posted here. I mean, seriously folks, I show screenshots taken from my (big) monitor. It should be clear that the image I'm working with isn't 400 pixels wide, eh? OK, on with the show."
posted by ShutterBun at 7:22 PM on April 12, 2011 [1 favorite]
I t
ho
ug
ht i
t w
as p
rett
y
coo
l!
posted by haricotvert at 8:03 PM on April 12, 2011 [4 favorites]
ho
ug
ht i
t w
as p
rett
y
coo
l!
posted by haricotvert at 8:03 PM on April 12, 2011 [4 favorites]
For those of you with an ios device, there's an app called Genius Scan that does exactly this. The results are somewhat limited by the quality of your camera, obviously, but it does a really nice job in general. It's a lifesaver for capturing notes on a blackboard.
posted by Captain Cardanthian! at 8:11 PM on April 12, 2011 [1 favorite]
posted by Captain Cardanthian! at 8:11 PM on April 12, 2011 [1 favorite]
Well, except that the picture on my wall that I take an oblique image of probably isn't up on flickr.
An image that has different content around the borders than the finished manipulated product. Because it was shot from a different angle. An image that also looks a lot cleaner than the serviceable manipulated image.
They didn't pretend to do this. It's not faked. This is, seriously, really basic stuff that makes perfect sense with a high resolution source pic being used to generate a not-so-high resolution manipulated pic. Anybody bothering to fake this by photoshopping up fake simple photoshoppery and editing the background details on the faked result to obscure that fakery would be wasting a lot of time on a non-prank; it'd be like photoshopping a fake image of a balloon static-clinging to someone's head and being like OMG GRAVITY DEFIED.
posted by cortex at 8:15 PM on April 12, 2011
An image that has different content around the borders than the finished manipulated product. Because it was shot from a different angle. An image that also looks a lot cleaner than the serviceable manipulated image.
They didn't pretend to do this. It's not faked. This is, seriously, really basic stuff that makes perfect sense with a high resolution source pic being used to generate a not-so-high resolution manipulated pic. Anybody bothering to fake this by photoshopping up fake simple photoshoppery and editing the background details on the faked result to obscure that fakery would be wasting a lot of time on a non-prank; it'd be like photoshopping a fake image of a balloon static-clinging to someone's head and being like OMG GRAVITY DEFIED.
posted by cortex at 8:15 PM on April 12, 2011
Computer, give me a picture of Oyster smiling.
posted by Meatbomb at 8:26 PM on April 12, 2011 [1 favorite]
posted by Meatbomb at 8:26 PM on April 12, 2011 [1 favorite]
I can think of a bunch of reasons why someone would post a fake photoshop tutorial to their blog, not least of which is that it still gets eyeballs for their ads almost two years after they posted it.
posted by crunchland at 9:05 PM on April 12, 2011
posted by crunchland at 9:05 PM on April 12, 2011
There's a much simpler way to get Photoshop to flatten a perspective-distorted rectangle. If you put the Crop tool in "perspective" mode, you can just drag the corners onto the corners of your quadrilateral, and hit Enter. Done in one step. If you also want some of the background, dragging an edge will move it in perspective, keeping it parallel with the imagined rectangle in perspective space.
Oh, and do fix your barrel/pincushion distortion before, not after, the perspective warp.
posted by otherthings_ at 9:34 PM on April 12, 2011
Oh, and do fix your barrel/pincushion distortion before, not after, the perspective warp.
posted by otherthings_ at 9:34 PM on April 12, 2011
Faking it would be more difficult and time-consuming than actually doing it. It would make no sense to fake it. To do so would be, if anything, Kaufmanesque, and not one of Andy's particularly good bits at that.
With no offense meant to Mitheral or the blogger, this just really isn't a notable thing to accomplish in Photoshop. It's really basic stuff. If the technique is new to you, cool, you have learned a new thing. But it's not a hoax; I've done the exact same thing more than once, it's tinker toys.
posted by cortex at 9:36 PM on April 12, 2011 [1 favorite]
With no offense meant to Mitheral or the blogger, this just really isn't a notable thing to accomplish in Photoshop. It's really basic stuff. If the technique is new to you, cool, you have learned a new thing. But it's not a hoax; I've done the exact same thing more than once, it's tinker toys.
posted by cortex at 9:36 PM on April 12, 2011 [1 favorite]
Please give us an example of the most amazing thing you can think of that can be done with photoshop. Unless it's being able to get a fingerprint of the reflection of a hand in a licence plate holder acron nut; I'd perfer to remain jaded about that piece of hollywood.I think the single coolest feature in Photoshop over the past 5 years is Adobe's hire/acquisition of Shai Avidan's "seam carving" technology (which they called "Content Aware Scaling") that allows you to delete exes from photos, make yourself taller, get that ugly guy out of the beach photo, etc.
http://en.wikipedia.org/wiki/Seam_carving
posted by Señor Pantalones at 9:45 PM on April 12, 2011
"WHY AM I NOT BEING DAZZLED PROPERLY"
dude dazzler is marvel
posted by Eideteker at 9:49 PM on April 12, 2011 [2 favorites]
dude dazzler is marvel
posted by Eideteker at 9:49 PM on April 12, 2011 [2 favorites]
I'm old school. I'd go and stand in front of it, if I was that interested in seeing it. I might make a tutorial about this technique.
posted by quarsan at 10:26 PM on April 12, 2011
posted by quarsan at 10:26 PM on April 12, 2011
With no offense meant to Mitheral or the blogger, this just really isn't a notable thing to accomplish in Photoshop. It's really basic stuff. If the technique is new to you, cool, you have learned a new thing. But it's not a hoax; I've done the exact same thing more than once, it's tinker toys.
I think that's the point. This is a fucking amazing thing to be able to do, but it's the fact that Photoshop makes it trivially easy to do once you know how to do it is what's so impressive.
Don't say "Feh, you can do this with two clicks." Say "Holy shit, with just two clicks, you can do this." There's no need to make what is magical merely banal.
posted by kafziel at 12:30 AM on April 13, 2011 [1 favorite]
I think that's the point. This is a fucking amazing thing to be able to do, but it's the fact that Photoshop makes it trivially easy to do once you know how to do it is what's so impressive.
Don't say "Feh, you can do this with two clicks." Say "Holy shit, with just two clicks, you can do this." There's no need to make what is magical merely banal.
posted by kafziel at 12:30 AM on April 13, 2011 [1 favorite]
Except this isn't magical in the least-- the pixels were always there, in the high resolution source image. In fact, all he really did was throw away vertical pixels to bring the aspect ratio back to normal. If you want to see magical, look at, for instance, blind deconvolution.
posted by Pyry at 1:39 AM on April 13, 2011 [2 favorites]
posted by Pyry at 1:39 AM on April 13, 2011 [2 favorites]
I'm a Photoshop tyro myself. Just getting started, really. Twenty minutes ago, I didn't know how to do this. Now I do. And I think that's pretty cool.
Haters gonna hate.
posted by Faint of Butt at 2:49 AM on April 13, 2011
Haters gonna hate.
posted by Faint of Butt at 2:49 AM on April 13, 2011
I'm not saying that it isn't possible to do anamorphic imagery tricks in photoshop, Cortex. I'm saying that with the picture he started with, I don't believe he got the picture he ended up with. The same way that I don't think it's possible for the people on CSI to read a license plate off the reflection on a car bumper from a security cam 4 blocks away. I just don't believe that the original image has enough sharpness on the poster to give him the crisp and sharp results. I'm inclined to believe that somewhere towards the end, he replaced the shabby version with the head on version, because no one would be interested in a blog post that said "hey, lookit this blurry photo I made."
posted by crunchland at 5:50 AM on April 13, 2011
posted by crunchland at 5:50 AM on April 13, 2011
Metafilter: I have no idea what you're (apparently) angry about.
posted by Grangousier at 5:53 AM on April 13, 2011
posted by Grangousier at 5:53 AM on April 13, 2011
Pyry- I'm glad someone's doing that kind of work, because I've been wanting to develop something similar and it shows that I might not be completely off my rocker for thinking this is possible.
Let's say you're holding a camera, and your hand is unsteady. You take a picture, and while the sensor is exposed to light, your hand (and thus the camera) undergoes some motion. For the sake of simplicity, we'll assume that the camera's motion can be described as having a translation component and a rotation component, neither of which are varying over the course of the camera's exposure (the faster the exposure is, the more accurately this will approximate the reality; this doesn't handle linear acceleration or oscillating motions).
Under those assumptions, there is a limited space of possible non-uniform* "blurs" that could be applied to the "true" image to arrive at the image that your camera reports. We want to search the space of blurs to find the blur that, when used as input to a (fast) deblurring algorithm, would produce the sharpest output image. Then we want to apply a slow, higher-quality deblurring algorithm using the source image and the blur we found.
*These non-uniform blurs can be approximated by a vector field, where the direction at any given point in the image is the velocity of the corresponding point on the camera's sensor. This representation is more intuitive than it sounds; in Photoshop, the "zoom" and "spin" blur dialog boxes show representations of such vector fields. From a "true" image, we can arrive at a "blurred" image by, for every pixel, doing a weighted average of the colors you encounter as you start from that pixel and move along the vector field for a few steps.
Reversing that sort of process would be tricky. The "fast" method would probably boil down to dividing the image up into chunks and using normal 2D deconvolution on those chunks. The slow method might form "paths" around the image by marching along the proposed vector field (except in the reverse direction), then doing a 1D deconvolution on each path. Or if you have a ton of free computer time and memory, just make a big-ass system of equations representing the proposed set of weighted sums.
posted by a snickering nuthatch at 6:25 AM on April 13, 2011
Let's say you're holding a camera, and your hand is unsteady. You take a picture, and while the sensor is exposed to light, your hand (and thus the camera) undergoes some motion. For the sake of simplicity, we'll assume that the camera's motion can be described as having a translation component and a rotation component, neither of which are varying over the course of the camera's exposure (the faster the exposure is, the more accurately this will approximate the reality; this doesn't handle linear acceleration or oscillating motions).
Under those assumptions, there is a limited space of possible non-uniform* "blurs" that could be applied to the "true" image to arrive at the image that your camera reports. We want to search the space of blurs to find the blur that, when used as input to a (fast) deblurring algorithm, would produce the sharpest output image. Then we want to apply a slow, higher-quality deblurring algorithm using the source image and the blur we found.
*These non-uniform blurs can be approximated by a vector field, where the direction at any given point in the image is the velocity of the corresponding point on the camera's sensor. This representation is more intuitive than it sounds; in Photoshop, the "zoom" and "spin" blur dialog boxes show representations of such vector fields. From a "true" image, we can arrive at a "blurred" image by, for every pixel, doing a weighted average of the colors you encounter as you start from that pixel and move along the vector field for a few steps.
Reversing that sort of process would be tricky. The "fast" method would probably boil down to dividing the image up into chunks and using normal 2D deconvolution on those chunks. The slow method might form "paths" around the image by marching along the proposed vector field (except in the reverse direction), then doing a 1D deconvolution on each path. Or if you have a ton of free computer time and memory, just make a big-ass system of equations representing the proposed set of weighted sums.
posted by a snickering nuthatch at 6:25 AM on April 13, 2011
This thread makes me want to be a better person at using photoshop.
posted by yeti at 6:49 AM on April 13, 2011 [1 favorite]
posted by yeti at 6:49 AM on April 13, 2011 [1 favorite]
I don't think I could recover this from this
While fixing perspective is trivial, I still struggle with levels/curves/contrast/etc... That resulting picture from the original is quite impressive to me.
posted by yeti at 6:51 AM on April 13, 2011
While fixing perspective is trivial, I still struggle with levels/curves/contrast/etc... That resulting picture from the original is quite impressive to me.
posted by yeti at 6:51 AM on April 13, 2011
That resulting picture from the original is quite impressive to me.
And that was produced with just a mild tweak to the default 'enhance' in iphoto, and 'auto-fix' in picnik.
posted by nomisxid at 7:30 AM on April 13, 2011
And that was produced with just a mild tweak to the default 'enhance' in iphoto, and 'auto-fix' in picnik.
posted by nomisxid at 7:30 AM on April 13, 2011
It strikes me that the OMG-amazing-modern-miracle part here is the existence of pretty easy, pretty affordable 8 megapixel or better cameras to provide the starting images. (Contributing good-luck factor in this exact case is that the subject of the photo was a cartoon character, hence you're working with large areas of flat color with sharp, contrasty edges. Without that, the sharpening pass wouldn't get you much.)
posted by jfuller at 7:31 AM on April 13, 2011
posted by jfuller at 7:31 AM on April 13, 2011
> I just don't believe that the original image has enough sharpness on the poster to give him the crisp and sharp results.
I, otoh, am a believer. I grabbed the final image in the tutorial (the one that contains both the purported perspective-corrected version and the purported head-on version and looked at them in xnview without distractions. Zoomed in some also, 200%--300%.
Considering especially the dark outlines around the areas of flat color, it seems to me that the original edge-on photo would have better data for capturing these in the case of lines that were horizontal from the photographer's POV and worse data for lines that were vertical, so that the reconstituted dark lines would be more successful (darker, crisper) for horizontal lines and less successful (less dark, less saturated, fuzzier) for vertical ones.
Looking at, say, the "wrist strengthener" (actually the shrunk-down little superhero vambrace on her left arm) that she's holding up. The curved top and bottom outlines of this piece (pretty much horizontal) in the two versions are very similar, while the straight vertical line separating the vambrace's internal color flat and the, uh, rays of power I guess, are quite different. In the claimed straight-on version this line is very much like the others in darkness and definition. In the claimed perspective-corrected version it's different, much paler and less defined. In short, exactly the difference I would expect to see between versions if the tutorial writer were telling the truth.
posted by jfuller at 8:07 AM on April 13, 2011
I, otoh, am a believer. I grabbed the final image in the tutorial (the one that contains both the purported perspective-corrected version and the purported head-on version and looked at them in xnview without distractions. Zoomed in some also, 200%--300%.
Considering especially the dark outlines around the areas of flat color, it seems to me that the original edge-on photo would have better data for capturing these in the case of lines that were horizontal from the photographer's POV and worse data for lines that were vertical, so that the reconstituted dark lines would be more successful (darker, crisper) for horizontal lines and less successful (less dark, less saturated, fuzzier) for vertical ones.
Looking at, say, the "wrist strengthener" (actually the shrunk-down little superhero vambrace on her left arm) that she's holding up. The curved top and bottom outlines of this piece (pretty much horizontal) in the two versions are very similar, while the straight vertical line separating the vambrace's internal color flat and the, uh, rays of power I guess, are quite different. In the claimed straight-on version this line is very much like the others in darkness and definition. In the claimed perspective-corrected version it's different, much paler and less defined. In short, exactly the difference I would expect to see between versions if the tutorial writer were telling the truth.
posted by jfuller at 8:07 AM on April 13, 2011
I just don't believe that the original image has enough sharpness on the poster to give him the crisp and sharp results.
The original image is, data-wise, much larger than the finished product. All it has to be is about as sharp along the horizontal as the output (and that's setting aside the nice pick-me-up the output gets thanks to the fact that the bold lines in the painting are line art that cleans up easier than a more complex color texture situation would).
And when you consider the approximate relative size of the original and the output, that's really dang plausible. The original at 25% vs. the final at 100% needs rectifying.
Note that my quickie inset here isn't even a fair treatment since we're comparing the cleaned up and sharpened final output against a crappy 60K lofi render of the original rather than the a proportionally high-quality version of same. But hopefully this at least puts it in context: the trick here is not pulling more horizontal data out of thin air than was originally available, but throwing away a whole bunch of vertical data in the process of correcting the orientation problems.
posted by cortex at 8:44 AM on April 13, 2011
The original image is, data-wise, much larger than the finished product. All it has to be is about as sharp along the horizontal as the output (and that's setting aside the nice pick-me-up the output gets thanks to the fact that the bold lines in the painting are line art that cleans up easier than a more complex color texture situation would).
And when you consider the approximate relative size of the original and the output, that's really dang plausible. The original at 25% vs. the final at 100% needs rectifying.
Note that my quickie inset here isn't even a fair treatment since we're comparing the cleaned up and sharpened final output against a crappy 60K lofi render of the original rather than the a proportionally high-quality version of same. But hopefully this at least puts it in context: the trick here is not pulling more horizontal data out of thin air than was originally available, but throwing away a whole bunch of vertical data in the process of correcting the orientation problems.
posted by cortex at 8:44 AM on April 13, 2011
While those are fun they are really using photoshop as a drawing tool. What I liked about the poster straightening was that it was pretty well a mechanical process to generate the output that isn't significantly different than reality.
IE: Those with talent can photoshop their minds eye even when their minds eye is reality, the poster thing appears at first glance to be pulling reality out of thin air.
posted by Mitheral at 9:46 AM on April 13, 2011
IE: Those with talent can photoshop their minds eye even when their minds eye is reality, the poster thing appears at first glance to be pulling reality out of thin air.
posted by Mitheral at 9:46 AM on April 13, 2011
Seeing is believing.*
*that photoshop can make some awesome images
posted by owtytrof at 9:49 AM on April 13, 2011
*that photoshop can make some awesome images
posted by owtytrof at 9:49 AM on April 13, 2011
I am trying to be impressed here. Perhaps this is something one would have to be a geek to appreciate.
Nope, Decani. Geek, and photoshop user. Figured out how to correct distortion and perspective on my own years ago - it's very useful for flash photography of shiny posters, scrolls, or flat objects behind glass.
First, you take the photo from an extreme angle, so the flash burst isn't visible in the shot.
Then, you correct distortion.
Then, you correct perspective.
The subject is well-lit and correctly scaled.
Pretty trivial. What am I missing here?
Next up: an amazingly talented photoshopper demonstrates how to replace the text on protest signs - with jokes!
posted by IAmBroom at 10:30 AM on April 13, 2011 [1 favorite]
Nope, Decani. Geek, and photoshop user. Figured out how to correct distortion and perspective on my own years ago - it's very useful for flash photography of shiny posters, scrolls, or flat objects behind glass.
First, you take the photo from an extreme angle, so the flash burst isn't visible in the shot.
Then, you correct distortion.
Then, you correct perspective.
The subject is well-lit and correctly scaled.
Pretty trivial. What am I missing here?
Next up: an amazingly talented photoshopper demonstrates how to replace the text on protest signs - with jokes!
posted by IAmBroom at 10:30 AM on April 13, 2011 [1 favorite]
This thread is the sort of conversation you'd have between a cave man who has seen an electric light bulb for the first time and a dude who's just switched it on. It's magic for the caveman, it's pretty quotidian for the dude.
Now the cavemen are trying the convince the dudes that it's magic and technology is amazing etc. and the dudes are all 'meh'. What can you say? Except don't tell anyone about livetrace.
posted by doobiedoo at 12:54 PM on April 13, 2011 [2 favorites]
Now the cavemen are trying the convince the dudes that it's magic and technology is amazing etc. and the dudes are all 'meh'. What can you say? Except don't tell anyone about livetrace.
posted by doobiedoo at 12:54 PM on April 13, 2011 [2 favorites]
« Older The Zippo Car | Soda > Slander & Lies Newer »
This thread has been archived and is closed to new comments
OK, I don't get it.
posted by ShutterBun at 5:19 PM on April 12, 2011 [2 favorites]