What's new
By:
Photoshop Gurus Forum

Welcome to Photoshop Gurus forum. Register a free account today to become a member! It's completely free. Once signed in, you'll enjoy an ad-free experience and be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

Underwater Photo Correction


Hi Bgood - Language has been a bit of problem in this thread, but I'm pretty sure the OP doesn't want to get rid of the light rays. If anything, I think he wants the light rays to stay at their current intensity or even be enhanced, but he really wants to get rid of the sensor noise that has been partially filtered by the raw rendering and JPG modules in his camera's SW.

Meeting both of these requirements simultaneously is the problem -- I tried several different commercial noise reduction software packages, and if it reduced the sensor noise by any appreciable amount, it also smoothed out the finest structure in the light rays. FYI, I tried (a) the NR facilities in ACR, (b) Neat Image, (c) Topaz DeNoise, plus a couple of public domain NR packages I have on my system. In the past, these have always worked better than Gaussian blur or similar DIY approaches.

Why don't you give it a shot and see what happens.

Cheers,

T
 
Tom Mann said:
Hi Bgood - Language has been a bit of problem in this thread, but I'm pretty sure the OP doesn't want to get rid of the light rays. If anything, I think he wants the light rays to stay at their current intensity or even be enhanced, but he really wants to get rid of the sensor noise that has been partially filtered by the raw rendering and JPG modules in his camera's SW.

Meeting both of these requirements simultaneously is the problem -- I tried several different commercial noise reduction software packages, and if it reduced the sensor noise by any appreciable amount, it also smoothed out the finest structure in the light rays. FYI, I tried (a) the NR facilities in ACR, (b) Neat Image, (c) Topaz DeNoise, plus a couple of public domain NR packages I have on my system. In the past, these have always worked better than Gaussian blur or similar DIY approaches.

Why don't you give it a shot and see what happens.

Cheers,

T
Click to expand...

My suggestion was based on my understanding of the request in the OP but I can see where this could well be a misunderstanding. Since I do not work much with photographs (hardly at all except to take photos of some of the pieces I have produced), I have not had many issues with noise and the few times I did I just used the clone tool to deal with it.

Thanks

Barry
 
Tom Mann said:
Hi Bgood - Language has been a bit of problem in this thread, but I'm pretty sure the OP doesn't want to get rid of the light rays. If anything, I think he wants the light rays to stay at their current intensity or even be enhanced, but he really wants to get rid of the sensor noise that has been partially filtered by the raw rendering and JPG modules in his camera's SW.

Meeting both of these requirements simultaneously is the problem -- I tried several different commercial noise reduction software packages, and if it reduced the sensor noise by any appreciable amount, it also smoothed out the finest structure in the light rays. FYI, I tried (a) the NR facilities in ACR, (b) Neat Image, (c) Topaz DeNoise, plus a couple of public domain NR packages I have on my system. In the past, these have always worked better than Gaussian blur or similar DIY approaches.

Why don't you give it a shot and see what happens.

Cheers,

T
Click to expand...

First of let me say thank a lot, really thank,
so glad to have found such good forum with so experienced members!
Thank a lot Tom Mann and bgoodman4 for such detailed answers and time that you Tom spent try processing my photo and help me.

Language has been a bit of problem in this thread, you right and sorry, cause my poor written english! I'm Italian

Tom how did you get all those info about my camera and setting that photo was shooted with?
Adobe camera raw?
I know basically photoshop, i know basically my way around, never give a try to camera raw cause i dont have a device capable of shooting in raw.
However Panasonic tz7 as a consumer, poket camera, is excellent, and you comments agrre with that, it shoots also very good video.

Can you suggest please a photo camera, a brand and a model capable of shoot in raw, tel me the brand please, do not make exp as you did before cause i did not understand much, i dont need a very professional machine cause i dont know much about photography however i would like to upgrade to a more professional device capable of shoot in raw.
I can order an underwater housing made for that particular model if the suggested one dont come with a commercial waterproof housing.

What about the commercial filters to reduce noise that you tried:
Neat image, it costs 40 $, 80$ with support for 16-32 bit image
Topaz DeNoise 80 $
which one of the 2 do you suggest i buy?

You tried to edit my photo and, if i understand well the final results where not so good, you where able to reduce grain however it ends up delating the nice sun light effects too, right?
Reduce grain and maintain at the same time the sun ray effects is the matter, meeting both of these requirements simultaneously is the problem.
Thank a lot!

 
Talking with you guys is gold to me,
i'm learning a lot so here another photo:

original_photo_2.jpg

again original dimension is 3900 px width, i reduce to 3500 px (constrain proportions), the maximum allowed.

Here no sun light effect to retain however as i apply the adjustment levels i end up increasing digital grain considerably.
May i wonder if i do something wrong,
basically i apply the tecnique that everybody use to correct underwater pictures where green or blue is the predominant color, in you tube there are many video that basically show the same tecnique:
adjustment levels that i apply:
exposure
chanel mixer
photo filter
levels
sometime i add vibrance too.

May the order in wich i apply the levels with is wrong?
Thank a lot
 
All those photos where obviously shooted to the same subject at the same moment


One more example,

original_photo_4.jpg

"original_photo_4", The Original one

correct_photo_4.jpg

"correct_photo_4", The Photo After manipulation

As you can see here results are much more good and again, i apply the same tecnique, the exact same adjustment layers mentioned before plus, once adjustments layers where applied, i merge all i a single one, "merge visible", duplicate the single layer that i end up with and set it to mutiply bland mode with 30% of opacity

Thank!
 
Hi Lanna - Just a few quick responses for the moment, more later.

a) How I got all the info about your camera: There are many programs that will do this. The best known is "EXIF tool" and "EXIF GUI". There are others that are less capable, but much more convenient because they are browser plugins. "FxIF data" is a nice one for Firefox. Just Google {EXIF metadata utilities}.

b) re the noise reduction plugins I used: I would recommend that you first work on the source of the problem (eg, get a new camera), not think about tools that should only be used as a last resort when trying to fix existing problems.

c) re a new camera: Probably the least costly type of camera that will greatly improve your current problems with digital noise, lack of RAW output, etc. would be one of the APS sensor (ie, small) DSLRs - "DX" series cameras in Nikon terminology. "FX" bodies are their much more expensive, full-frame models.

I'm most familiar with Nikon equipment, so I would recommend something like the d7100 (~$1200), but there are Canon and other equivalents. Given your use, I would purchase it without the usual kit zoom lens, and instead, pick up an inexpensive 28 mm f/2.8 fixed focus AF lens, If you are willing to spend double this amount, then you can move up to full-frame, professional bodies which will improve the digital noise even more.

d) Record all your images in RAW+jpg, but process only the RAW files using ACR, then PS. Process all files at 16 bits per channel bit depth from the moment they leave the camera till the very last moment when you have to convert them to 8 bit per channel ( 8 bpc) JPGs for display on the web. Do NOT shoot in any wide-gamut color space like ProPhoto. Do everything in plane old sRGB. This will further reduce noise and banding problems.


Later ....

Tom
 
Thank Tom!
I understand perfectly the essence of your suggestions, buy a proper camera first, this must be the first steep and i'm going to buy it for sure, i'm going to check your devices suggestions on the web. Telling the true i'm a little bit concern about shooting underwater with "big in size device" cause at the end of the day i'm spearfishing, camera or videocam (midland) is attached to my gun, impossible to attach very big device plus the underwater housing, anyway this is my problem nothing to do with the subject here in the forum.
No matter i'm going to use it underwater while spearfishing, i'm going to buy a more professional camera.
In the main time about pictures that already have, i have, very nice to me, photos and videos about all my spearfishing trips around the world, my web site below, nothing commercial:
http://www.extremespearfishing-worldtravel.com/
i woould like publish them in my web site, edit them in a proper way when edit is needed, etc, unfortunately all those photos are jepg. So about those pictures what do you suggest, what about buying the plug in that you mentioned before?
I check about Topaz, it seems to be very good stuff for rasonable prices, Topaz DeNoise and Topaz DeJPEG, plus in you tube there are many and very detailed tutorials about those plug ins.
i can try with those plug in for the photos that i already have and for the future i'm going to follow your suggestions, buy a more professional device that allows raw output.
Thank a lot
 
Last edited:
Tom, could I trouble you for your thoughts on the following, more just wondering and thinking out loud than actually a suggestion.

If the water were selected for and copied to another layer and then blurred slightly (or possibly more than slightly), and then the new water layers opacity were reduced, could that possibly help hide the artifacts you are talking about? Also, I know very little about masks as I have never used them in my work (probably should though) but from reading many of the threads on this forum perhaps something along this line might work better than the 2nd layer thing I just outlined.
 
Hi Bgood - There have been tens of thousands of academic papers, web articles, different pieces of software, etc. all written on noise reduction in digital images. The bottom line is that there is always a tradeoff between noise reduction and blurring of desired features. The method(s) you proposed are very reasonable, and were among the 1st ever tried, but much better methods have since been developed.

Unfortunately, I have to run to work, so I don't have time to go into this in more detail at the moment, but will try to get back to you late tonight on this.

Regards,

Tom
 
Tom Mann said:
Hi Bgood - There have been tens of thousands of academic papers, web articles, different pieces of software, etc. all written on noise reduction in digital images. The bottom line is that there is always a tradeoff between noise reduction and blurring of desired features. The method(s) you proposed are very reasonable, and were among the 1st ever tried, but much better methods have since been developed.

Unfortunately, I have to run to work, so I don't have time to go into this in more detail at the moment, but will try to get back to you late tonight on this.

Regards,

Tom
Click to expand...

Thanks Tom, much appreciated.
 
Tom Mann said:
Hi Bgood - ...I tried several different commercial noise reduction software packages, and if it reduced the sensor noise by any appreciable amount, it also smoothed out the finest structure in the light rays. FYI, I tried (a) the NR facilities in ACR, (b) Neat Image, (c) Topaz DeNoise, plus a couple of public domain NR packages I have on my system. In the past, these have always worked better than Gaussian blur or similar DIY approaches.

Why don't you give it a shot and see what happens.

Cheers,

T
Click to expand...

Tom,
what about topaz denoise and dejpeg? I watched tutorials about, in you tube planty of Topaz plu ins detailed one, they seem to be very good plug ins generally speaking.
I ask the following question cause you gave a try with this plug in, if i undertood well the final results where not so good however you mentioned only Topaz denoise and suggested to give it a try anyway.
About all my hold photos, all shooted by my panasonic tz7 so jpeg output, do you think buying those 2 plugins can solve at list in part my problems, do they allow me to edit my photos sorting up the matter of this thread? You gave it a try already so can you please describe telling me what do you think about your test and more in depth generally speaking?
Thank a lot
 
@Bgood - What makes one NR method better than another is not the ability of one to smooth out variations in the image better than the other, but rather, the ability of one to correctly identify what is real detail (ie, not noise) and to deal with those areas differently (eg, smooth them less, apply a modified algorithm in those regions, etc.).

The method(s) that you proposed have absolutely no "smarts" in them. They have no ability to identify real details (other than perhaps, manually, say, with a layer mask), so they simply apply the same degree and type of smoothing everywhere on the image. The result of this often is a kind of plastic look imparted to the image.

In contrast, more modern NR algorithms will look for edges or certain spatial frequency bands or certain types of texture (eg, the fractal-based NR algorithms), etc. and then reduce the smoothing of such features. This (a) allows one to crank up the smoothing of the rest of the image, and (b) gives a more realistic look.

HTH,

Tom M
 
Hi Lanna -


re: "... you mentioned only Topaz denoise and suggested to give it a try anyway. ..."

When I suggested to BGood that "he give it a try anyway", I was not suggesting that he try Topaz DeNoise or any of the algorithms that I tried. Rather, I was referring to the simple, do-it-yourself method(s) that he had proposed earlier such as Gaussian blur on a copy of the layer at partial opacity, etc.
---------

re: Topaz's "de JPG"

This program is designed to reduce one, very specific type of artifact, the 4x4 blocky artifacts that arise from too low of a quality factor when generating a JPG file (see this article for a nice intro to these types of artifacts). The images you have posted don't show any evidence of this problem. It might be there, but it's much less than the other artifacts you have in your image. JPG artifacts are almost always seen around high contrast edges. Your underwater images have few truly high contrast edges.
----------

re: purchasing Topaz's "de Noise" as a stop-gap measure till you get a camera with lower noise

You already own and are reasonably proficient at Photoshop CS6. ACR contains quite a good noise reduction algorithm (see attached). Before you purchase a 3rd-party plugin, I suggest you get familiar with ACR's NR capabilities. If you don't know how to have your JPGs automatically go through ACR, there was a thread on exactly this just yesterday.

It turns out that the type of images on which Topaz's "deNoise" does best is reasonably similar to the types of images on which ACR's NR algorithm does well. If you really want to try a different NR algorithm than the one in ACR, I suggest you try "Neat Image" instead. It nicely complements the other two algorithms.

HTH,

Tom M
 

Attachments

  • NR_in_ACR-screenshot.jpg
    NR_in_ACR-screenshot.jpg
    94.9 KB · Views: 21
Thank a lot Tom and every body,
i'm going to follow your suggestions.
some notes:
i mentioned Topaz dejpeg cause i have other photos taken by friends that definitely need that plug in, i did not mention that before, sorry!
I took in consideration Topaz DeNoise as first option cause, when i decide to buy software generally speaking, i check informations related, if that software has tutorials, etc, too. About denoise, it comes with a lot of tutorials, very a few about Neat Image. However i understood perfectly what you ment, Neat image is easy to use, dont need to watch many tutorials and, as you said, it has an algorithm that completes the one that i have already, photoshop and camera raw.
About camera raw, it works with raw, jpeg, and other output, i assumed wrongly that it was a software mainly for raw output, reading info about i understood that works equally good with jpeg output so, cause i never gave it a try, i'm going to learn as you suggested and try to solve my editing problem with those softwares
thank a lot!

One final question, sorry, it has nothing to do with this thread subject:
saving a jpeg for the web,
how much compression i must use? Usually i save at 100%, best quality as photoshop preset, is it wrong? In term of best quality, i dont concern much about file dimension, it is better save them at 100% or saving them at 80-90%, photoshop preset as high quality, etc, i end up with an output that is almost identical in quality and smaller file size to 100% best quality?
Thank again!
 
Hi Lanni -

The question of selecting reasonable values for the "quality" and "pixel dimensions" when saving JPGs often is asked. In a minute, I'll post a separate "sticky" thread in the photography section on this topic.

HTH,

Tom
 
Thank,
i ask about "save for web" quality cause i read very different comments, some tell to save at highest quality possible, while some other say that it is much better less quality than 100 %, waiting for your answer tom.

I begun to watch tutorials about camera raw and, as a beginner as i'm, i get confused so sorry about my stupid question:
about camera raw and jpeg, despite using raw with jpeg output for minor adjustment as in this case try to reduce noise, if i understood well, while working with jpeg, better to use photoshop am I right or not?
Sorry again for my beginner stupid question
Thank
 
Tom Mann said:
@Bgood - What makes one NR method better than another is not the ability of one to smooth out variations in the image better than the other, but rather, the ability of one to correctly identify what is real detail (ie, not noise) and to deal with those areas differently (eg, smooth them less, apply a modified algorithm in those regions, etc.).

The method(s) that you proposed have absolutely no "smarts" in them. They have no ability to identify real details (other than perhaps, manually, say, with a layer mask), so they simply apply the same degree and type of smoothing everywhere on the image. The result of this often is a kind of plastic look imparted to the image.

In contrast, more modern NR algorithms will look for edges or certain spatial frequency bands or certain types of texture (eg, the fractal-based NR algorithms), etc. and then reduce the smoothing of such features. This (a) allows one to crank up the smoothing of the rest of the image, and (b) gives a more realistic look.

HTH,

Tom M
Click to expand...

Understood, thanks again.
 
Back
Top