Pages: 1 ... 7 8 [9] 10 11 ... 16   Go Down

Author Topic: ZERO NOISE technique  (Read 401701 times)

docmaas

  • Newbie
  • *
  • Offline Offline
  • Posts: 34
ZERO NOISE technique
« Reply #160 on: November 07, 2007, 05:09:55 pm »

Welcome back.  I discovered this thread just after you left.  I hope when you write your program you will include the sigma x3f raws in it.  

I am sure there are many of us hanging on to this thread waiting for further news.

Best

Mike
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1843
    • http://www.guillermoluijk.com
ZERO NOISE technique
« Reply #161 on: November 07, 2007, 05:31:25 pm »

Quote
Welcome back.  I discovered this thread just after you left.  I hope when you write your program you will include the sigma x3f raws in it. 

I am sure there are many of us hanging on to this thread waiting for further news.

Best

Mike
[{POST_SNAPBACK}][/a]

Thank you. Actually since I am not developing the files myself but using [a href=\"http://www.guillermoluijk.com/tutorial/dcraw/index_en.htm]David Coffin's DCRAW[/url], the formats supported will be those supported by DCRAW, which is a wide list (at present I think is the only RAW developer which can deal with Nikon D300 RAW files hehe).

Is your camera one of these?
- Sigma SD9
- Sigma SD10
- Sigma SD14

They are supported.

I would like to write some code this weekend.
« Last Edit: November 07, 2007, 05:32:55 pm by GLuijk »
Logged

docmaas

  • Newbie
  • *
  • Offline Offline
  • Posts: 34
ZERO NOISE technique
« Reply #162 on: November 07, 2007, 08:10:46 pm »

I have all three but the sd10 is broken.  Great to hear.  

thanks,

Mike

Quote
Thank you. Actually since I am not developing the files myself but using David Coffin's DCRAW, the formats supported will be those supported by DCRAW, which is a wide list (at present I think is the only RAW developer which can deal with Nikon D300 RAW files hehe).

Is your camera one of these?
- Sigma SD9
- Sigma SD10
- Sigma SD14

They are supported.

I would like to write some code this weekend.
[a href=\"index.php?act=findpost&pid=151164\"][{POST_SNAPBACK}][/a]
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10143
ZERO NOISE technique
« Reply #163 on: November 07, 2007, 10:19:26 pm »

Quote
Well Ray, I have to say what Diapositivo said is right: my routine just generates an image free of noise in the shadows, but with the same exposure, bright, contrast,... and everything as the least exposed shot of the set used. Therefore it is very dark and is the user's choice to choose the best way to lift the shadows where and as much as he wants. I wanted a routine that does not modify the original image's parameters in any way.

[a href=\"index.php?act=findpost&pid=151132\"][{POST_SNAPBACK}][/a]

Guillermo,
I can see that your method could have some advantages with regard to ease of processing. If one has difficulty in getting the desired result, as many people seem to, often with merged to HDR images, it's not clear whether the procedure itself makes it impossible to get the right result or whether the user is not skilled enough to get the desired result with that particular procedure.

The 'Layer Mask' method in the LL tutorial gives you a blended image with two layers, one representing the shadows and the other the highlights. One has to play around with these two layers, adjusting the levels, whatever, until the image looks right. Sometimes the image might never look right, possibly due to the fact that the initial exposures were too far apart, or due to one or both of the the RAW images requiring a different EC adjustment on conversion.

I'm all in favour of simplifying techniques for the user. I thought the shadow/highlight control first intoduced in PSCS was a really useful improvement to Photoshop, as is the new contrast/brightness tool in CS3 which protects the highlights.

In your example image of what looks like a deserted car park, the initial image is so dark, it would hardly matter if there were noise in the shadows. In lightening the image to demonstrate how noise-free the shadows are, you've lost detail in the highlights.

If you had used the 'layer mask' method in the LL tutorial, you would have been able to lighten that image by adjusting only the 'dark' layer without losing any highlight detail.  
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1843
    • http://www.guillermoluijk.com
ZERO NOISE technique
« Reply #164 on: November 08, 2007, 02:29:45 am »

Of course Ray, I just applied a global lifting curve to show that all the information is available and noise-free, both in the shadows and in the highlights. But for a proper edition is the user who has to play with the layer mask of the curves or any other tool available to get the desired 'manual tone mapping'. What I wanted to avoid is that a program (like Photomatix and etc...) decides for me where and how much the shadows are to be lifted obtaining unrealistic results.

So the result provided is simply what you would get with a noise-free camera that allows to capture the whole dynamic range in the shadows when setting the exposure to preserve the highlights of the scene, and it is you who decide now your preferred edition method.

A more adequate edition of that image obtaining the perceived atmosphere of the place could be (I think I only used a curve without layer mask but taking care of preserving the light hole):



BTW that strange place is not a car park hehe, it is a very weird commemorial that was built for the 200 victims killed in Madrid by Al Qaeda. The light coming from the roof is natural light taken from outside and the place is deliberately poorly lighted resulting in a high dynamic range scene to our eyes.
« Last Edit: November 08, 2007, 02:52:01 am by GLuijk »
Logged

MichaelEzra

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1112
    • https://www.michaelezra.com
ZERO NOISE technique
« Reply #165 on: November 08, 2007, 09:14:27 am »

Guillermo,

This is a wonderful method of obtaining HDR image. The tonemapping operation to get the LDR image is the next step:)! One could also use a free Picturenaut from hdrlabs.com

I will try to send you the NEF files from the Nikon Coolscan 8000 scanner tonight.
« Last Edit: November 08, 2007, 09:14:47 am by MichaelEzra »
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10143
ZERO NOISE technique
« Reply #166 on: November 08, 2007, 01:47:23 pm »

Quote
So the result provided is simply what you would get with a noise-free camera that allows to capture the whole dynamic range in the shadows when setting the exposure to preserve the highlights of the scene, and it is you who decide now your preferred edition method.
[a href=\"index.php?act=findpost&pid=151244\"][{POST_SNAPBACK}][/a]

Guillermo,
The 'Layer/mask' method outlined in the LL tutorial does have its difficulties, namely halos that sometimes occur at abrupt transitions between dark and light areas. I was initially interested in your method because it seemed to solve this problem.

However, HDR in CS3 seems to work just fine with only 2 images and with no halo problems.

The fact is, in this situation of blending 2 images, one cannot get less noise than the least noisy of the two images, which is the overexposed image or the image correctly exposed for the shadows.

It occurred to me that the relatively clean shadows in the overexposed image might get contaminated with noise from the underexposed image during the blending process to HDR. If this were the case, then your method would have a clear advantage because, as you've explained it, the pixel values are unchanged. Your blended image consists of either pixels from one image or the other, whereas the other blending methods probably create new values that don't exist in either image prior to blending.

I think there's probably no better test of blending methods than to shoot a scene from inside a room of the view through a window. The contrast is extremely high and the transition from light to shade around the window frame is very sharp.

So to settle this matter in my own mind, ie. whether the blended image is more noisy in the shadows than the unblended overexposed image, I took a few shots of the window in my rather squalid apartment which I'm renting for $10 a day on a monthly basis with unlimited broadband included. (Saving up for a 5D MkII, see   ).

[attachment=3753:attachment]  [attachment=3754:attachment]  [attachment=3755:attachment]

[attachment=3756:attachment]  [attachment=3757:attachment]

The final image compares 100% crops of the shadows in the overexposed image prior to merging to HDR, with the same area after merging to HDR and lightening.

To my mind, noise levels are about the same in each crop.
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1843
    • http://www.guillermoluijk.com
ZERO NOISE technique
« Reply #167 on: November 08, 2007, 03:11:57 pm »

mmm this is very interesting Ray, I always thought PS HDR was progressive in the entire image area, but your test shows that at least in some areas it takes just the information from the overexposed shot (which is the optimum). In fact I saw a video of PS HDR merging of different shots on a street scene with walking people, and they all appeared as ghosts (semi transparent people), which meant progressive blending.

I only have CS2 and trying PS HDR with it I got this non-optimum results in some areas (see background wall below), so as strange artifacts in the highlights due to interference of the overexposed shot (see lamp below).
Maybe CS3 has fixed these problems, could you please try this image with your CS3 to find out? these are the RAW files:

SUB.CR2
SOBRE.CR2


(you can see a small slightly noisy area bottom left in the ZERO NOISE image, where the switch took place)



Entire image:


Regards

PS: BTW I always feel a bit embarrashed for showing such large pictures, but don't know how to set the 'click to enlarge' thumbnails from them. How should I do it?
« Last Edit: November 08, 2007, 03:18:49 pm by GLuijk »
Logged

Quentin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1222
    • Quentin on Facebook
ZERO NOISE technique
« Reply #168 on: November 08, 2007, 04:09:44 pm »

I hope Michael might review the software here on LL when it is ready.  In any event, my credit card stands ready to pay a fair price when its released.

Quentin
Logged
Quentin Bargate, ARPS, Author, Arbitrato

jule

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 738
    • http://www.juliestephenson.net
ZERO NOISE technique
« Reply #169 on: November 08, 2007, 09:00:22 pm »

Quote
Guillermo,


So to settle this matter in my own mind, ie. whether the blended image is more noisy in the shadows than the unblended overexposed image, I took a few shots of the window in my rather squalid apartment which I'm renting for $10 a day on a monthly basis with unlimited broadband included. (Saving up for a 5D MkII, see   ).

[a href=\"index.php?act=findpost&pid=151346\"][{POST_SNAPBACK}][/a]

I see you have air-con Ray...hope it works!

Julie
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10143
ZERO NOISE technique
« Reply #170 on: November 09, 2007, 12:47:59 am »

Quote
I see you have air-con Ray...hope it works!

Julie
[a href=\"index.php?act=findpost&pid=151421\"][{POST_SNAPBACK}][/a]

Julie,
The air-con works just fine. It's quite new but I don't need it just at the moment. It's quite cool in Chiang Mai at the monent.

Broadband in Australia seems to start off at A$29.95 per month for a very limited download of 200MB. Thereafter, the speed either drops down to dial-up 56kbps or one pays through the nose for additional megabytes. Also one usually has to sign a year's contract, which means it's going to cost more to take up the inevitable better offer that comes along in 3 months or 6 month's time.

It's no wonder John Howard is going to lose the next election. This government has been very slack in encouraging the development of broadband.

By contrast, unlimited high speed broadband in this hotel/apartment complex I'm staying at, in a relatively underdeveloped country, costs just A$27 per month for a minimum of 1 month. The room, about 40 sq m, is costing A$7.20 a day, plus broadband of 89 cents a day plus a small amount for electricity and water makes a total of around A$10 a day.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10143
ZERO NOISE technique
« Reply #171 on: November 09, 2007, 01:48:49 am »

Quote
PS: BTW I always feel a bit embarrashed for showing such large pictures, but don't know how to set the 'click to enlarge' thumbnails from them. How should I do it?
[a href=\"index.php?act=findpost&pid=151364\"][{POST_SNAPBACK}][/a]

Guillermo,
It seems to be automatic when you download images from your computer's hard drive, rather than a website, using the 'Browse and Add Attachment' facility at the bottom of the 'reply' page.

But your method is better because the full size images are instantly viewable.

I've downloaded your 2 RAW images at lightning speed and tried a few blending techniques. In the process I managed to redecorate your apartment. Hope you like the result   . (Sorry about the spill on the floor and sofa, but we artistic types are a bit sloppy. We need someone to clean up, after).

[attachment=3765:attachment]

Could it be you have some colorful, fluorescent undercoating on your walls which  only CS3 can pick up?  

Seriously, this result only occurs when I convert the images before merging to HDR. Merging the RAw images produces a fairly normal result but still not satisfactory. These exposures are about 4 stops apart. I can only assume that the 'merge to HDR' process needs an intermediate exposure to get the tonalities right. So your method, at least with these two examples, would appear to be a big improvement on PS's HDR.

But I was curious as to how the layer/mask method described in the LL tutorial would handle these 2 images.

Here are the results. First the two converted images prior to blending. The dark image has been lightened by +1 EC and the light image has been darkened by -1 EC during conversion.

[attachment=3766:attachment]  [attachment=3767:attachment]

When applying the gaussian blur of the B&W mask I used a radius setting of 100 pixels. LL recommends 40 pixels, but this is variable according to circumstances. The following image shows the resulting blend without further levels adjustment of the two layers.

[attachment=3768:attachment]

The balance of this image seems about right in accordance with what the eye would see but probably still a little too dark in the deepest shadows, if you want to be literal.

I tried to match your very dark image, which is not what the eye would see, and came up with the following, by adjusting levels and contrast in the individual layers.

[attachment=3769:attachment]

Now for some noise-free 100% crops.

[attachment=3770:attachment]  [attachment=3771:attachment]  [attachment=3772:attachment]

I should mention that none of these images have been sharpened, which is one reason for their looking so noise-free. I'll also make a general comment that your images with your blending method seem slightly more even in the tonality. On the other hand, I don't know how the original scene really appeared to the eye.
« Last Edit: November 09, 2007, 02:21:25 am by Ray »
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1843
    • http://www.guillermoluijk.com
ZERO NOISE technique
« Reply #172 on: November 09, 2007, 06:06:29 am »

A lot of information in a single post, step by step:

1. I am a bit confused about those green colours PS CS3 HDR provided you in my sitting room, is that the only result than can be achieved? I merged the images in CS2 HDR and found non-optimum results and artifacts, but do you  mean CS3 HDR has got worse?  

2. I must admit I have not read LL method, but looking at your explanations I guess it consists of merging 2 versions of the same image with some difference in exposure, making use of a gaussian blur to make the blending progressive. Right?
Your result is great and natural, but I don't agree with this step: "The dark image has been lightened by +1 EC" since it means you are blowing 1 complete f-stop of information in the highlights.
However it will surely work also by leaving the dark image as is (so no loss of information), and applying a -2EV correction to the light image and then the LL method. Bright could be then controlled using a curve which preserves detail in the highlight while exposure correction doesn't.

3. Regarding the final crops, I wonder how the left area of the black cushion results in the LL method. The right side you showed seems to show a bit more noise (specially colour noise) than with my method (which would mean LL method took some info from the dark image even for the black cushion), but I wonder if you can show us the left side.

This is what I get in the right side of the cushion:


And this was the left side:


PS: BTW I see you reescaled a bit down the image. For noise comparisions, if the image has necessarily to be reescaled it is VERY IMPORTANT to perform a nearest neighbout reescaling (in the Spanish version of PS it is called 'By Aproximation', and is the first option of PS reescaling methods, following next Bilinear and Bicubic).
Nearest neighbour reescaling just selects some unmodified pixels from the original image, therefore it preserves intact the signal to noise ratio as can be seen on a 100% crop, while any interpolation method (bicubic, bilinear,...) reduces noise thanks to pixel averaging.
Also integer 50%, 33.3%, 25%, 20%,... reescalings are recommended when using nearest neighbour to avoid aliasing artifacts.

Regards
« Last Edit: November 09, 2007, 06:08:55 am by GLuijk »
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10143
ZERO NOISE technique
« Reply #173 on: November 09, 2007, 11:19:14 pm »

Quote
1. I am a bit confused about those green colours PS CS3 HDR provided you in my sitting room, is that the only result than can be achieved? I merged the images in CS2 HDR and found non-optimum results and artifacts, but do you  mean CS3 HDR has got worse? 

Guillermo,
When using TIFFs for merge to HDR, a warning message appears to the effect that it's better to use RAW images. The conversions I used for HDR were the same as I used for the LL layer/mask technique. Obviously Photoshop's HDR program didn't like my settings   .

I tried again using the unconverted RAW images as advised, but there was still a hint of that discoloration. I've got no idea why. When I reconverted the RAW images with EC at zero and all the other settings at zero, the result was much improved as you can see in the comparison crops below.

This seems very strange to me. A warning message advising one to use RAW images, but a clearly better result when using TIFFs.

[attachment=3781:attachment]

Below is the full image after merging to HDR using TIFF conversions, followed by the lightened version using the shadows/highlight tool and a crop of the dark cushion lightened further. I'm not sure how these images will appear on a calibrated desktop monitor. I'm using a laptop which has been only roughly calibrated with Adobe Gamma.

[attachment=3782:attachment]  [attachment=3783:attachment]  [attachment=3784:attachment]

Quote
2. I must admit I have not read LL method, but looking at your explanations I guess it consists of merging 2 versions of the same image with some difference in exposure, making use of a gaussian blur to make the blending progressive. Right?
Your result is great and natural, but I don't agree with this step: "The dark image has been lightened by +1 EC" since it means you are blowing 1 complete f-stop of information in the highlights.
However it will surely work also by leaving the dark image as is (so no loss of information), and applying a -2EV correction to the light image and then the LL method. Bright could be then controlled using a curve which preserves detail in the highlight while exposure correction doesn't.

I'm not sure what's going on here under the hood. As I mentioned, this layer/mask technique tends to produce halos. To avoid getting halos, or to reduce the severity of the halos, it seems to be necessary to make EC adjustments at the conversion stage to minimise the tonal differences between the light image and the dark image, hence the light image -1 EC and the dark image +1 EC. I did not get the impression I was blowing highlights by using +1 EC with the dark image, perhaps because I also used the 'recover highlights' slider in ACR. Essentially the darker image has large specral highlights which always remain blown whatever the setting (ie. the inside of the lamp).

Below is the initial blended image from the layer/mask procedure using the same TIFFs I used for the above 'merge to HDR', ie. conversions with zero settings.

[attachment=3785:attachment]

It's difficult to now get rid of these halos with further adjustment. It's better to avoid them at the start. This is why I am interested in your method. However, HDR is producing a good result, no?

Quote
PS: BTW I see you reescaled a bit down the image. For noise comparisions, if the image has necessarily to be reescaled it is VERY IMPORTANT to perform a nearest neighbout reescaling (in the Spanish version of PS it is called 'By Aproximation', and is the first option of PS reescaling methods, following next Bilinear and Bicubic).
Nearest neighbour reescaling just selects some unmodified pixels from the original image, therefore it preserves intact the signal to noise ratio as can be seen on a 100% crop, while any interpolation method (bicubic, bilinear,...) reduces noise thanks to pixel averaging.
Also integer 50%, 33.3%, 25%, 20%,... reescalings are recommended when using nearest neighbour to avoid aliasing artifacts.

I didn't realise that. All the downsized full images have been rescaled using bucubic. The small 100% crops have not been rescaled but I did make changes from 240ppi to 72ppi whilst maintaining the same file size, ie. increasing the dimensions so the file size remains the same. Is that still rescaling?
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10143
ZERO NOISE technique
« Reply #174 on: November 10, 2007, 02:06:46 am »

Guillermo,
I've tried various conversion settings to try and find out what it is that HDR does not like in some of my converted tiff images. It appears to be the contrast of the dark image. It's not sufficient to have the contrast setting at zero in ACR. It must be at minus 50 to avoid that discoloration. However, it doesn't seem to make much difference if the contrast of the light image is set at zero or -50.

Nevertheless, the fact remains that I'm able to get a better result by making my own conversions instead of letting the HDR program do them.
Logged

simonkit

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 67
    • Landscape Photography UK.com
ZERO NOISE technique
« Reply #175 on: November 24, 2007, 07:18:27 pm »

Some fascinating stuff on here - thanks all for the excellent info I'll certainly be giving the suggestions a try

 simon
Logged
[url=http://www.landscapephotographyuk.c

wolfy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 93
ZERO NOISE technique
« Reply #176 on: November 25, 2007, 01:48:42 pm »

Quote
Jan, just because my post followed yours (I didn't quote yours either) doesn't mean my remarks were addressed solely to you: they were addressed to all those who when offered something for free were ungrateful and wanted more.
Peter
[a href=\"index.php?act=findpost&pid=139871\"][{POST_SNAPBACK}][/a]


Great thread and probably great product,...one which this Mac user will appreciate and pay for, if/when available. And would use via :"Virtual PC"-like application if available and effective.

Some support for Jani's comments:

Someone offers free shoes, ..mentioning a certain size. You ask if YOUR (different) size is available.

How is this indicative of un-gratefulness. It is a straightforward and reasonable QUESTION, no?

Why characterize the asking negatively? It only lends credence to the anti-Mac-user hostility mentioned by Jani.

BTW, Mac-only users (as opposed to users of both, such as yourself), are by definition excluded form the group which was  been "offered" something. The question was intended to find-out whether they would be included at some future time.

I have not seen anything posted in this thread indicating any non-appreciation of the OP's work, ...at most some respectful and reasonable questioning.

"Eye of the beholder", etc.  
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1843
    • http://www.guillermoluijk.com
ZERO NOISE technique
« Reply #177 on: March 24, 2008, 11:40:04 pm »

Beta v0.9 version of Zero Noise, the program to automate this technique.

Hi all it's said it's better late than never. I took some time these days to finally develop the first beta version of the blending program to minimise noise and expand DR. There are still some things to improve and add, but at the moment its fully functional and performs nicely in the tests.

I will offer it for download very soon from my site with a tutorial on how to make use of it.

Meanwhile have a look at this micro-tutorial with an example. Sorry it's in Spanish (by now) but you can follow the images to find out.


Micro tutorial:

1. OPEN RAW FILES Se indica con la opción '...' el directorio donde están los RAW a fusionar.

Con elegir uno basta, el programa leerá los demás mostrando la imagen seleccionada y la lista total de RAWs (lo he autolimitado a 10 RAWs, pero usar más de 4 deja de tener sentido en cualquier aplicación. Con 3 haciendo bracketing 0,+2,+4 como en este ejemplo los resultados son buenísimos):




2. WHITE BALANCE Se ajusta el balance de blancos. Como aún no está la opción de temperatura/matiz, y los multiplicadores lineales de DCRAW pueden ser poco intuitivos para el prueba/error, he introducido la posibilidad de seleccionar un parque rectangular o circular sobre la imagen que será balanceado en blancos. Este parche lo dibuja el usuario con tan solo clickear sobre la imagen:



si se prefiere un parche circular porque la zona de interés para balancear se parece más basta pulsar el botón que indica un cuadrado:




De las 2 posibilidades me quedo con la primera por resultar más natural:




3. RAW DEVELOPMENT Una vez tenemos el balance de blancos deseado, no hay más que pulsar 'Develop' y el programa invocará a DCRAW para que revele los RAW. Para ver el progreso de DCRAW es bueno desactivar el checkbox llamado 'Hide MS-DOS' que hay en la parte de abajo:




4. BLENDING El paso anterior ha generado un archivo .tiff por cada RAW suministrado. Ahora solo hay que pulsar la opción 'Blend' (mezcla) y el programa los fusionará en una imagen final con el ruido minimizado pues tomará cada píxel del RAW menos ruidoso. Mostrará al final del proceso las exposiciones relativas entre tomas; este paso es muy importante, el programa calcula numéricamente cuáles fueron, ignorando los EXIF que pueden ser totalmente engañosos (subiré un ejemplo donde se ve esto muy bien):



En un bracketing -2,0,+2 podemos ver que la separación en EV entre la primera y segunda toma no fue de 2EV sino menos. Caso de haber usado 2EV en la mezcla se habrían notado las transiciones entre zonas.


5. MANUAL TONE MAPPING El resultado será un .tiff lineal que se podrá leer en PS con tan solo asignarle una versión lineal del espacio de color que se usó como salida. De este enlace se pueden descargar versiones lineales de sRGB y de Adobe RGB.

Una vez cargada y asignada al espacio de color, con tan solo convertir la imagen al espacio de color destino (que puede ser el mismo en que la generamos, solo que ya no será una versión lineal), ésta se deslinealizará y quedará lista para ser editada como cualquier imagen revelada normal.

Esta imagen final parecerá bastante subexpuesta y sin contraste. La imagen NO ESTÁ SUBEXPUESTA, Y NO TIENE BAJO EL CONTRASTE, los tiene tal cual salen del RAW sin procesar. Es solo que acostumbrados a que ACR y demás reveladores apliquen por su cuenta curvas y ajustes de brillo éstas parecerán sosas.

Basta dos curvas (una de levantamiento de sombras y otra de contraste) para tener una imagen final correcta de alto rango dinámico, con las altas luces preservadas y bajo ruido en las sombras, sin haber hecho ninguna reducción de ruido que nos pudiera hacer perder texturas:




Comparando la toma menos expuesta del conjunto inicial (la única en que no aparecía quemado el exterior de la ventana) con la toma resultante:

« Last Edit: March 24, 2008, 11:41:08 pm by GLuijk »
Logged

docmaas

  • Newbie
  • *
  • Offline Offline
  • Posts: 34
ZERO NOISE technique
« Reply #178 on: March 24, 2008, 11:51:57 pm »

Great News.  Looking forward to the release.

Best,

Mike

Quote
Beta v0.9 version of Zero Noise, the program to automate this technique.

Hi all it's said it's better late than never. I took some time these days to finally develop the first beta version of the blending program to minimise noise and expand DR. There are still some things to improve and add, but at the moment its fully functional and performs nicely in the tests.

I will offer it for download very soon from my site with a tutorial on how to make use of it.

Meanwhile have a look at this micro-tutorial with an example. Sorry it's in Spanish (by now) but you can follow the images to find out.
Micro tutorial:

1. OPEN RAW FILES Se indica con la opción '...' el directorio donde están los RAW a fusionar.

Con elegir uno basta, el programa leerá los demás mostrando la imagen seleccionada y la lista total de RAWs (lo he autolimitado a 10 RAWs, pero usar más de 4 deja de tener sentido en cualquier aplicación. Con 3 haciendo bracketing 0,+2,+4 como en este ejemplo los resultados son buenísimos):


2. WHITE BALANCE Se ajusta el balance de blancos. Como aún no está la opción de temperatura/matiz, y los multiplicadores lineales de DCRAW pueden ser poco intuitivos para el prueba/error, he introducido la posibilidad de seleccionar un parque rectangular o circular sobre la imagen que será balanceado en blancos. Este parche lo dibuja el usuario con tan solo clickear sobre la imagen:



si se prefiere un parche circular porque la zona de interés para balancear se parece más basta pulsar el botón que indica un cuadrado:


De las 2 posibilidades me quedo con la primera por resultar más natural:


3. RAW DEVELOPMENT Una vez tenemos el balance de blancos deseado, no hay más que pulsar 'Develop' y el programa invocará a DCRAW para que revele los RAW. Para ver el progreso de DCRAW es bueno desactivar el checkbox llamado 'Hide MS-DOS' que hay en la parte de abajo:


4. BLENDING El paso anterior ha generado un archivo .tiff por cada RAW suministrado. Ahora solo hay que pulsar la opción 'Blend' (mezcla) y el programa los fusionará en una imagen final con el ruido minimizado pues tomará cada píxel del RAW menos ruidoso. Mostrará al final del proceso las exposiciones relativas entre tomas; este paso es muy importante, el programa calcula numéricamente cuáles fueron, ignorando los EXIF que pueden ser totalmente engañosos (subiré un ejemplo donde se ve esto muy bien):



En un bracketing -2,0,+2 podemos ver que la separación en EV entre la primera y segunda toma no fue de 2EV sino menos. Caso de haber usado 2EV en la mezcla se habrían notado las transiciones entre zonas.
5. MANUAL TONE MAPPING El resultado será un .tiff lineal que se podrá leer en PS con tan solo asignarle una versión lineal del espacio de color que se usó como salida. De este enlace se pueden descargar versiones lineales de sRGB y de Adobe RGB.

Una vez cargada y asignada al espacio de color, con tan solo convertir la imagen al espacio de color destino (que puede ser el mismo en que la generamos, solo que ya no será una versión lineal), ésta se deslinealizará y quedará lista para ser editada como cualquier imagen revelada normal.

Esta imagen final parecerá bastante subexpuesta y sin contraste. La imagen NO ESTÁ SUBEXPUESTA, Y NO TIENE BAJO EL CONTRASTE, los tiene tal cual salen del RAW sin procesar. Es solo que acostumbrados a que ACR y demás reveladores apliquen por su cuenta curvas y ajustes de brillo éstas parecerán sosas.

Basta dos curvas (una de levantamiento de sombras y otra de contraste) para tener una imagen final correcta de alto rango dinámico, con las altas luces preservadas y bajo ruido en las sombras, sin haber hecho ninguna reducción de ruido que nos pudiera hacer perder texturas:


Comparando la toma menos expuesta del conjunto inicial (la única en que no aparecía quemado el exterior de la ventana) con la toma resultante:


[a href=\"index.php?act=findpost&pid=184042\"][{POST_SNAPBACK}][/a]
Logged

bernie west

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 321
    • Wild Photo Australia
ZERO NOISE technique
« Reply #179 on: March 25, 2008, 05:46:58 am »

good work Guillermo.  Can't wait to try it.
Logged
Pages: 1 ... 7 8 [9] 10 11 ... 16   Go Up