ZERO NOISE photography using Photosop

[Guillermo Luijk]

Some time ago I tried a technique to eliminate (severely reduce actually) noise in digital images through multiexposure: http://www.guillermoluijk.com/article/nonoise/index_en.htm.
I am going to write a program to do it automatically from the RAW files with expanded features, but another forum member (Joan Trujillo) has posted recently his method to perform the same operation in Photoshop. The application of his method is incredibly simple, conceptually robust and of course effective.
 
This is the link to his PS method: http://jtrujillo.net/qpix/
'Yet another method to reduce noise with two exposures' by Joan Trujillo.

 
I recommend you give it a try, it's a really good system to blend 2 images achieving noise reduction and dynamic range expansion.
 
 
This is a test I did with two pictures +4EV apart:
 
Adjust layers and original less exposed image saving highlights without any adjustment.

[Sunesha]

Some time ago I tried a technique to eliminate (severely reduce actually) noise in digital images through multiexposure: http://www.guillermoluijk.com/article/nonoise/index_en.htm.
I am going to write a program to do it automatically from the RAW files with expanded features, but another forum member (Joan Trujillo) has posted recently his method to perform the same operation in Photoshop. The application of his method is incredibly simple, conceptually robust and of course effective.
 
This is the link to his PS method: http://jtrujillo.net/qpix/
'Yet another method to reduce noise with two exposures' by Joan Trujillo.

 
I recommend you give it a try, it's a really good system to blend 2 images achieving noise reduction and dynamic range expansion.
 
 
This is a test I did with two pictures +4EV apart:
 

[a href=\"index.php?act=findpost&pid=161975\"][{POST_SNAPBACK}][/a]

Thanks man,

That is great.  I Must try this when I come home. I am über grateful for the tutorial. After reading your other posts I tried myself but never figured out the technique

Cheers,
Daniel

[John Sheehy]

Some time ago I tried a technique to eliminate (severely reduce actually) noise in digital images through multiexposure: [a href=\"index.php?act=findpost&pid=161975\"][{POST_SNAPBACK}][/a]

When you posted about this last time, I thought that perhaps you may have come to regret using the phrase "zero noise".  You are still using it now, but your software doesn't result in zero noise, so why call it that?  It merely defers the shadow noise into deeper shadows.  There is noise at every tonal level, regardless of how you blend different exposures.

[Guillermo Luijk]

When you posted about this last time, I thought that perhaps you may have come to regret using the phrase "zero noise".  You are still using it now, but your software doesn't result in zero noise, so why call it that?  It merely defers the shadow noise into deeper shadows.  There is noise at every tonal level, regardless of how you blend different exposures.
[a href=\"index.php?act=findpost&pid=161980\"][{POST_SNAPBACK}][/a]

Marketing tactiques  

No, seriously, I know it is not strictly zero noise John, but with the adequate overexposure you can make noise so low that will be impossible to detect to the eye, at any f-stop. Then I think that, if not 100% scientifically correct, is a quite fair claim. For most photographers, if they were asked about noise in these images, they would simply say there is no noise.

But I understand your point very well.

PS: anyway you can see the author took care to be more honest than me, so if the claim is wrong, I am the only responsible.

[tandlh]

Nothing new here.  See Michael's column on Digital Blending and drill down to "The Layer Mask".  This has been on this site for quite some time now.  

http://luminous-landscape.com/tutorials/di...-blending.shtml

Juan does do a good job explaining it again though.

[John Sheehy]

Nothing new here.  See Michael's column on Digital Blending and drill down to "The Layer Mask".  This has been on this site for quite some time now. 

http://luminous-landscape.com/tutorials/di...-blending.shtml

Juan does do a good job explaining it again though.
[a href=\"index.php?act=findpost&pid=162038\"][{POST_SNAPBACK}][/a]

Well, what is relatively unique here is that it is done at the RAW level, within a modified version of DCRAW.

[Guillermo Luijk]

it is done at the RAW level, within a modified version of DCRAW.

John, have you ever come across someone who has developed a floating point version of DCRAW? so that more precision is kept if gamma or other processes are applied before final 16-bit rounding?

Another question, what would you think of overexposing a lot a picture, develop it with DCRAW (linear 16-bit), and then read it with a program that deals with the 16-bit levels and will apply a exposure correction down in floating point, so no rounding, and blend it with a less exposed shot that preserves the highlights. And then we apply gamma 2.2 before final rounding the output? that output would be read from PS as already gamma corrected of course.

Do you think that would produce noticeable improvement in the shadows' tonal richness? I am going to try that in the next days but I wonder if it will be worth the effort.

Regards.

[jtrujillo]

Nothing new here.  See Michael's column on Digital Blending and drill down to "The Layer Mask".  This has been on this site for quite some time now. 

http://luminous-landscape.com/tutorials/di...-blending.shtml

Juan does do a good job explaining it again though.
[{POST_SNAPBACK}][/a]

Hi,

I am Juan, the author of that article Guillermo talks about ([a href=\"http://jtrujillo.net/qpix]http://jtrujillo.net/qpix[/url])

I would like to clarify some details, not to refute your reasoning (especially when you say I did a good job hehe) but to point to the relevant basement of what we are doing and of my method. I must say that I read Michael’s article a long time ago (the stuff on the site is an essential part of my background :-).

I know didn’t discover anything and don’t claim it. But I’d like to analyze the strong point of the method that I am proposing (maybe this is new, I haven’t seen it anywhere as is here). I feel we should work to differentiate it from other methods just to grasp its potential)

Aside, notice that we correct the exposure in the overexposed shot in the RAW program to put it at the same level as the well-exposed shot (the one exposed for the lights)

In short what we do is placing on top of the well exposed shot the one overexposed-and-corrected-in-RAW. We need something to select from that top layer only the relevant information –discard the clipped data- Everything else is contributing noise-almost-free data to the result.

Well, the core of this thing is in the layer mask, how easily is built and how much control gives us to fine tune the result.

The way to construct that layer is simply using ***the overexposed shot, not exposure-corrected and inverted***

Why is so special?
1. Any black pixel in that layer mask corresponds to a highlight clipped in the overexposed shot and with no more work we know positively will not slip though the result.

2. By applying levels to that mask you control the gradation of the blend and can set thresholds to that gradation simply moving the black and the white sliders.

3. You can work ghosting very easily: Painting the mask black or white you tell the system to use one or other layer exclusively

Regards,
Juan

[jtrujillo]

I though I'd put another sample, this with 100% crops of the result with and whitouth the processing

Final image:


100% crop:


Same crop but without the layer that does the correction