|
Maybe I don't know enough about how everything works, but it sounds like one of those "convert your old VHSs into super high-quality DVDs!!" products. How does it restore data that wasn't captured to begin with? Interpolation only goes so far, and it's not like your camera is hiding parity data in EXIF or something. I am interested in seeing the results though, so please post some when you can.
|
#
?
Dec 2, 2010 08:04
|
|
|
|
| # ? Jun 8, 2024 02:34 |
|
|
Yeah, go to the website and hover over the license plate image. http://www.topazlabs.com/infocus/#4 I call bullshit.
|
|
#
?
Dec 2, 2010 08:07
|
|
|
ExecuDork posted:I used Recuva when I accidentally deleted about 200 photos from my SD card
|
|
#
?
Dec 2, 2010 08:51
|
|
|
I downloaded the free trial and dicked around with it. I used the ruler tool to find the angle of the blur.  was what I got from  Kinda cool I guess, but the result will always look shittacular.
|
|
#
?
Dec 2, 2010 09:01
|
|
|
Haggins posted:It looks like it's time to stop deleting out of focus shots. Topaz Labs just came out with a new product called Infocus, which supposedly fixes out of focus images. Looking at the examples on the site it seems pretty impressive. What the fuuuucckkk? How does this black magic work? It can't just be simple sharpening but... Downloading now, trip report tomorrow.
|
|
#
?
Dec 2, 2010 09:36
|
|
|
dakana posted:I downloaded the free trial and dicked around with it. That said, I don't believe in miracles; I'm sure they picked the best possible examples to use on their site. I'm gonna download for myself and see what it can do.
|
|
#
?
Dec 2, 2010 16:03
|
|
|
The above example is really impressive, you cant even read the plate in the original. Some of the examples on the page just make it look like "Topaz unsharp mask" though
|
|
#
?
Dec 2, 2010 16:30
|
|
|
Cyberbob posted:This might help Thanks. I also found this on the same site that seems to focus on making realistic images. http://photoshoptutorials.ws/photoshop-tutorials/photo-retouching/hdr-tone-mapping-with-layers-in-photoshop.html
|
|
#
?
Dec 2, 2010 17:17
|
|
|
Trip report: this is a great plugin for fixing images that you want to display at 220x150px. Or, like Dakana, you're a PI taking pictures of politician's license plates as they speed away from a hotel rendezvous with a stripper. I tried it on three images where I missed focus: some just barely, some by a lot. Didn't get anything that looked better than unsharp mask or high pass filter, and they were a hell of a lot noisier. Again, I wasn't expecting miracles, I just wanted something potentially usable. Original, 100% crop.  High pass at 9px, overlay layer at 50%:  Unsharp mask at 50, 1.0, 0:  Topaz on the "common" setting:  Now considering that their examples are shrunk to something like 25%, their originals must look pretty hosed up if you saw them at 100%. I honestly can't see how they even got them looking as good as they did shrunk down, but I'll bet they're a noisy mess if you saw them at full size. Maybe you can save a few photos from the reject bin, but you still wouldn't have anything great.
|
|
#
?
Dec 2, 2010 17:36
|
|
|
jackpot posted:Again, I wasn't expecting miracles, I just wanted something potentially usable. I spent like 2 minutes on it and got a lot better result with InFocus than you did:  I masked out the background too to reduce the noise in it, and I think if you practice with it more it'll become a lot simpler and more effective. I think it is a really good deal for $30, but I wouldn't pay the full $70 for it.
|
|
#
?
Dec 3, 2010 05:11
|
|
|
Yeah, you definitely did better with it than I did, that's pretty good. Like you said though, $70 is drat steep for something I can do 90% as well with unsharp mask.
|
|
#
?
Dec 3, 2010 16:57
|
|
|
You can't create information out of what isn't there, this just looks like some kind of modified Smart Sharpen + contrast.
|
|
#
?
Dec 3, 2010 21:04
|
|
|
brad industry posted:You can't create information out of what isn't there, this just looks like some kind of modified Smart Sharpen + contrast. The license plate test was pretty interesting though. If the program understands the path the camera took while the shutter was open (like if someone bumped the tripod), it seems within the realm of possibility that the program could figure out which pixels should go where.
|
|
#
?
Dec 3, 2010 21:40
|
|
|
That is pretty much what it looks like it's doing I think. It arranges them based on the amount of movement and angle according to edge contrast (basically sharpness, you can probably get the same result doing it manually in Smart Sharpen). It looks shittacular because once it does the rearranging it has no information to fill in the gaps. edit: I think the reason that ocean one worked is because water/wet things are super contrasty and have really abrupt changes in value usually limited to one channel, so that is way easier for it to work with than something that has smooth tones or transitions. brad industry fucked around with this message at 21:57 on Dec 3, 2010 |
|
#
?
Dec 3, 2010 21:53
|
|
|
Its like that pedo guy who got busted because he used the whirlpool/twisty function in photoshop to hide his face in pictures, and someone just ran it in reverse to identify him.
|
|
#
?
Dec 3, 2010 22:20
|
|
|
I while ago Microsoft published some information on doing this by using gyroscopes and some other bolt-on stuff. This is the before/after examples obviously, they look pretty good, but then, the source image wasn't that bad: 
|
|
#
?
Dec 4, 2010 00:35
|
|
|
mobby_6kl posted:I while ago Microsoft published some information on doing this by using gyroscopes and some other bolt-on stuff. This is the before/after examples obviously, they look pretty good, but then, the source image wasn't that bad: Isn't this just image stabilization but in reverse?
|
|
#
?
Dec 4, 2010 01:08
|
|
|
Hey guys, I'm writing some software for work and it involves setting camera parameters to take "nice" pictures. Does anyone know of a good background on how matrix and evaluative metering works? As in a little bit of an overview of the algorithms involved?
|
|
#
?
Dec 4, 2010 01:11
|
|
|
AIIAZNSK8ER posted:Isn't this just image stabilization but in reverse? It used the gyroscopes to embed camera movement in the exif data. With that data they used software to counter act it which is pretty cool. If i remember they knew the software was not great but if the information was recorded that software down the road could do a much better job.
|
|
#
?
Dec 4, 2010 01:25
|
|
|
Isn't there a point where fixing failure might be better served in just going outside and taking better pictures?
|
|
#
?
Dec 4, 2010 01:25
|
|
|
Paragon8 posted:Isn't there a point where fixing failure might be better served in just going outside and taking better pictures? Of course, but the research is still interesting. It could be adapted for use on things more interesting than fixing a lovely snapshot.. restoring old photos, improving the quality of security footage, finding out who really killed JFK, whatever. I'm not saying they'll be doing Blade Runner style enhancement, but improving the quality of the data that is there is a pretty interesting topic.
|
|
#
?
Dec 4, 2010 01:32
|
|
|
sund posted:Hey guys, I'm writing some software for work and it involves setting camera parameters to take "nice" pictures. Does anyone know of a good background on how matrix and evaluative metering works? As in a little bit of an overview of the algorithms involved? Not much is known officially about the algorithms being used. Some info on Nikon's old system can be found here.. beware it's from Ken : http://www.kenrockwell.com/nikon/matrix.htm My naive approach for a software implementation would be to segment the image into a grid and use each grid cell as an input node for a neural net, maybe add some center weighting. Then train the net with as many "nice" photos and associated EXIF info as you can find and hope that the actual results will be nice too.
|
|
#
?
Dec 4, 2010 01:51
|
|
|
If anyone is curious on how Infocus works, if you go to this video: http://www.youtube.com/watch?v=9_W_f4Ft7PE Then fast foward to the 1:39:30 mark, Ray Maxwell explains how it works. It's not sharpening and it's not something you can do in Photoshop.
|
|
#
?
Dec 4, 2010 05:01
|
|
|
Denoise seemed to work really well. Going to try it afterwards on a real noisemaker (Olympus E-420 @ ISO1600).
|
|
#
?
Dec 4, 2010 13:45
|
|
|
Quick and dirty test. No other edits apart from resizing to 800x600. Pack of smokes on the floor at ISO1600:  100% crop:  After 1 minute of sliding sliders in Denoise 5:  100% crop:  It does get a bit of that cartoony blur that comes with all noise reduction but given how little effort I put into it and the natural limits of turd polishing I have to say that's pretty good. You don't need to remove every bit of grain to make it usable anyway. Once you got some experience you could probably preserve much more detail while still removing most of the noise. I didn't use debanding by the way, but can see some banding issues now. That could possible contribute to the loss of detail in the black lines on the floor.
|
|
#
?
Dec 4, 2010 15:27
|
|
|
Ola posted:Quick and dirty test. No other edits apart from resizing to 800x600. Quick and dark test. I can see results but can't help but not care unless you want to expose your poo poo correctly.
|
|
#
?
Dec 4, 2010 21:01
|
|
|
benisntfunny posted:Quick and dark test. I can see results but can't help but not care unless you want to expose your poo poo correctly. My exposure wasn't on trial, noise removal was. I tried to get it as noisy as possible.
|
|
#
?
Dec 4, 2010 22:08
|
|
|
Ola posted:My exposure wasn't on trial, noise removal was. I tried to get it as noisy as possible. High ISO only works a drat on a correctly exposed image, so that kinda needs to be done first IMO.
|
|
#
?
Dec 5, 2010 11:45
|
|
|
Royking Dreper, bastard son of Don Roy King and Don Draper. (No really, what's that mean?)
|
|
#
?
Dec 5, 2010 12:07
|
|
|
Cyberbob posted:High ISO only works a drat on a correctly exposed image, so that kinda needs to be done first IMO. But I wasn't testing high ISO was I? Geez, is there a pretty-threshold for anything posted in this thread? I took a picture to maximize the noise and then tried to see what the app could do. It's not a pretty picture of a pack of cigarettes that's going on my Flickr, it's a picture of noise. Is this hard? I can take a pretty one afterwards if that makes you happy. pwn posted:Royking Dreper, bastard son of Don Roy King and Don Draper. (No really, what's that mean?) Smoking kills! It doesn't kill noise obviously.
|
|
#
?
Dec 5, 2010 12:12
|
|
|
I'm thinking of selling my 50mm 2.5 macro and getting some tubes until I can afford the 100mm. I have the 70-200mm f4 non is, the obligatory 17-50 and the 50mm 1.8. Am I going to miss having the 50mm 2.5? It is pretty sharp for more than just macro, and it was my only lens for the longest time. I have a bit of a soft spot for it, but since getting the 70-200 and 17-50, I rarely use it unless I'm doing macro stuff.
|
|
#
?
Dec 5, 2010 15:55
|
|
|
Ola posted:But I wasn't testing high ISO was I? Geez, is there a pretty-threshold for anything posted in this thread? I took a picture to maximize the noise and then tried to see what the app could do. It's not a pretty picture of a pack of cigarettes that's going on my Flickr, it's a picture of noise. Is this hard? I can take a pretty one afterwards if that makes you happy. Testing something like noise removal should be done in a real life example. Why would I care that you can remove noise from a mostly black lovely picture? That tells me absolutely nothing. There is not enough detail and color range to accurately judge if that's doing a good job or not.
|
|
#
?
Dec 5, 2010 21:01
|
|
|
Cyberbob posted:High ISO only works a drat on a correctly exposed image, so that kinda needs to be done first IMO. But an underexposed photo will result in a much noisier image. If you want to test noise reduction in a worst-case scenario then underexposing your photo is pretty valid. e: Beaten. e2: I don't know, I think it's a pretty decent test.
|
|
#
?
Dec 5, 2010 21:04
|
|
|
benisntfunny posted:Stop defending your poor test. While I agree that it's a pretty bad test overall, I do think that testing noise removal in a (slightly) underexposed image can indeed be a real world test. I know that I have found myself in situations where I have had no choice other than to underexpose slightly, otherwise there would be subject blur in the shot. e: bad test is a little harsh-- I think it might be more fair to say that it's not a super useful test. Noise in a web size 800x600 is negligible most of the time, there is no comparison against a properly exposed 1600ISO shot/overexposed 1600ISO shot pulled down, no sharpening done to the image (which would exacerbate the noise, but in a real shot, generous sharpening is usually done before posting/printing), etc. Shmoogy fucked around with this message at 21:12 on Dec 5, 2010 |
|
#
?
Dec 5, 2010 21:07
|
|
|
Shmoogy posted:e: bad test is a little harsh-- I don't think poor test is really that inaccurate of a statement though; I did not say bad test - though I guess those words are almost interchangeable. And while the initial image might be underexposed in real life the noise would likely occur from a person trying to bring it to the correct exposure. I guess moral of the story is... no one would ever use that image as is so it's not entirely useful. I appreciate the fact that someone is willing to do some real world testing but if they're going to do it wrong it's going to waste everyone's time and possibly give less knowledgeable people misconceptions.
|
|
#
?
Dec 5, 2010 21:28
|
|
|
benisntfunny posted:I don't think poor test is really that inaccurate of a statement though; I did not say bad test - though I guess those words are almost interchangeable. Apologies, I was the one who said 'bad test' and I felt it was a bit harsh. quote:And while the initial image might be underexposed in real life the noise would likely occur from a person trying to bring it to the correct exposure. quote:I guess moral of the story is... no one would ever use that image as is so it's not entirely useful. I appreciate the fact that someone is willing to do some real world testing but if they're going to do it wrong it's going to waste everyone's time and possibly give less knowledgeable people misconceptions.
|
|
#
?
Dec 5, 2010 21:48
|
|
|
Shmoogy posted:Agreed, that's why I felt that you were correct in stating that it was a poor test, and why my edit said that I felt it was not a very useful test. Oh sorry, it seems I just came in and got all defensive and confused myself ;D
|
|
#
?
Dec 5, 2010 21:51
|
|
|
benisntfunny posted:Stop defending your poor test. This is a pretty weird attitude. I wasn't doing you all a service, there was a discussion on the Topaz software and I got interested in the denoise thing. If you don't like my underexposed picture of a cigarette pack I dropped on the floor beneath my computer desk (EV -1.7 bitches) you are welcome to try it yourself, there's a free trial on that as well. Noise vs shake is a choice you make when you take pictures in poor lighting, perhaps one is a better tradeoff than the other when in such a situation. Maybe you got some idea of the software's ability to remove noise while retaining detail compared to other things you've tried, but I'm not worried about making it look pretty for you.
|
|
#
?
Dec 5, 2010 22:52
|
|
|
Ola posted:This is a pretty weird attitude. I wasn't doing you all a service, there was a discussion on the Topaz software and I got interested in the denoise thing. If you don't like my underexposed picture of a cigarette pack I dropped on the floor beneath my computer desk (EV -1.7 bitches) you are welcome to try it yourself, there's a free trial on that as well. Noise vs shake is a choice you make when you take pictures in poor lighting, perhaps one is a better tradeoff than the other when in such a situation. Maybe you got some idea of the software's ability to remove noise while retaining detail compared to other things you've tried, but I'm not worried about making it look pretty for you. Or you could have boosted the exposure, assuming it was a raw file, to a proper visible level, which is what a good low light/high noise image test would be.
|
|
#
?
Dec 5, 2010 23:33
|
|
|
|
| # ? Jun 8, 2024 02:34 |
|
|
How can I tell to what extent a photographer has used post-production to achieve the look of their photos? How do I know if a really good looking photo has only slightly been altered (if at all), when it was the photographer's skill with his or her camera settings that achieved the majority of the quality? As an example, look at jackpot's recent excellent thread here: http://forums.somethingawful.com/showthread.php?threadid=3368162 To what extent has post production been used for most of the photos? I'm still relatively new to photography (compared to the skill levels of most people here), and I'm still a far cry from achieving photos of this standard. I have the equipment for it, but not the skill. But like I say, I'm finding it difficult to learn from others when I'm not sure how much of what I'm looking at is artificial.
|
|
#
?
Dec 5, 2010 23:43
|
|
