Please please please! Help me to sort out my HDR from a single Raw image

[golclub]

Hi Guys

I took a raw image from my A720IS and then in Photomatrix used Single File Convesation where I converted the raw file into pseudo-HDR image and saved it as Radiance RGBE (.hdr) file and then opened and tonemapped it in Photomatrix. I attached the result and looks disastrous. Extremely noisy and dull. I tried millon other ways and results were the same or worse. I am interested only in working with one raw image as I want to take picture of moving object.

Guys, just beg you to help me to sort out the problem.

Thanks!

[coralnutz]

Do you have a picture of the original? Looks like it was extremely underexposed in order to up the exposure became very noise.

[JoYork]

If its any consolation a lot of single RAW pseudo-HDR files can look noisy even from a DSLR, and the sensor in those are much bigger.

What ISO were you using? I don't even attempt conversion if the ISO is anything higher than ISO 100.

You might be better off trying out a plugin such as Topaz Adjust which helps create a pseudo-HDR tonemapped look. It even includes a de-noise element within it.

[golclub]

Do you have a picture of the original? Looks like it was extremely underexposed in order to up the exposure became very noise.


I have an original and will gladly post it somewhere, but this forum does not allow to upload anything bigger than 150 MB

:(

[JoYork]

Try this:

http://www.upload.sc/

[golclub]

here is the raw file

http://rapidshare.com/files/187198408/CRW_7357.DNG.html

[canonloader]

Here is my attempt at converting your dng to an hdri and adjusting it in tone mapping in Photomatix... I finished it in CS3 to get rid of noise and resize.
http://www.picturelacrosse.com/test/hdr/parkinglot-012109sm.jpg

And here is the xmp file (http://www.picturelacrosse.com/test/hdr/parkinglot.zip). Put the xmp file into the presets folder in the Photomatix install folder on your hard drive, then open Photomatic, open your HDRi file in Photomatix, click the tone mapping button, then in the tool panel, at the bottom, use this xmp file to set the sliders to see what I did.

[golclub]

Conanloader, Big THANKS for doing this for me :D:D

I applied that preset and it looks cool, but colours I had were very different from what I can see in your post. I really do not understand why. Check my example after applying your preset on
http://rapidshare.com/files/187576101/CRW_7357parklo1t.jpg.html

Could you tell me please what optrions you used in Photoshop to get rid of noise?

My picture still is far from perfect in terms of noise, this is because the original file had to much noise. How should I take pictures so that there is no or little noise like in pictures I can see on flickr?

Thanks a lot !

[canonloader]

What version of Photoshop do you use? Your picture isn't that much different than mine, but I tweaked mine in CS3 after exporting as a tif from Photomatix. I then opened it to CS3 as a RAW so I could use the sliders to adjust light and color. I just added some saturation for the colors. Then saved as a jpg and reopened that in CS3 and cropped, resized, ran a NR action I have and then sharpened it back up a little. It all took a couple minutes. :)

Make sure the original is properly exposed to cut down on any noise to start with. It also helps to start with a big file dimension and then reduce it. If you crop and keep it big, you will get noise too. So many thing introduce noise when you process the file that you have to keep that to a minimum.

[JoYork]

I had a go with this image too, in different HDR converters as well as some Photoshop plugins.

I don't think it's really the best candidate for HDR or tonemapping. I can understand you wanting to give the image a bit of "pop", but you might have better results taking another photo when the light's different.

[golclub]

Make sure the original is properly exposed to cut down on any noise to start with...

...but you might have better results taking another photo when the light's different.


Thank you guys for helping to understand the process of making HDRs

So in order to avoid noise, the original RAW file should be properly exposed then, yeah? Because I am a novice in photography I would like to ask how to make picture better exposed?

Could also tell what is the best way of dealing with a single raw image? Can I just open it in Photomatrix and then tonemap it? Or do you think the best way is to save it with different exposures in some program, such as Raw Therapee for example, as TIFF 16 bit files and then merge then in Photomatrix? I used all ways and none of them was really satisfactory. You can see that on

http://www.flickr.com/photos/batyrinuk

Regards

[boomer3297]

Make sure the original is properly exposed to cut down on any noise to start with...

...but you might have better results taking another photo when the light's different.


Thank you guys for helping to understand the process of making HDRs

So in order to avoid noise, the original RAW file should be properly exposed then, yeah? Because I am a novice in photography I would like to ask how to make picture better exposed?

Could also tell what is the best way of dealing with a single raw image? Can I just open it in Photomatrix and then tonemap it? Or do you think the best way is to save it with different exposures in some program, such as Raw Therapee for example, as TIFF 16 bit files and then merge then in Photomatrix? I used all ways and none of them was really satisfactory. You can see that on

http://www.flickr.com/photos/batyrinuk

Regards

Ultimately the histogram should be exposed as far to the right as possible without clipping the highlights. As was said before noise is an inherent problem with single exposure HDR and is really only recommended on low contrast scenes. This is according to Ferrell Mc Collough's book "the complete guide to HDR digital photography".bw!

[JoYork]

Also bear in mind that this isn't really HDR at all. Your camera appears to have had enough dynamic range to capture the scene perfectly - what you're effectively doing is exaggerating the extremes, making the dark bits lighter and the light bits darker. As soon as you start making dark bits lighter you exaggerate the noise in the image.

To keep the noise levels down and produce some Photomatrix-friendly images you need to expose to the right as much as you can, keep the ISO as low as possible and...well, get lucky, basically. I've tonemapped single RAW files with varying degrees of success - some turn out great, more turn out average or worse than what I started with.

To expose to the right you'll preferably need to shoot manual mode, but you could also use exposure comp or try various tricks with your metering and exposure locking. Check the histogram religiously and try not to clip too many highlights (particularly the sky).

[golclub]

Thank you boomer and JoYork!

I will do what you recommended. As I am a novice to photography, all these terms are pretty vague to me. Could you just tell me in laymen terms what clipping highlights and keep to the right is?

Sorry, if I sounded silly:-)

[kirkt]

An attempt - not sure you need to tonemap with an HDR-like operator, but I ran the DNG through FDRTools and then did some PS work. It could probably use a little more local contrast, but you get the idea. I tried to go for a more toned down, natural look - but, of course, I have never seen this place so I am just guessing at what the scene might look like. If you just operate on the image with ACR and/or PS, you can probably control the noise better, as tonemapping operators that are used in HDR apps often accentuate local contrast (noise).

http://kirkt.smugmug.com/photos/460254953_CEvaj-L.jpg

Here is a 100% crop of your unprocessed DNG:

http://kirkt.smugmug.com/photos/460277682_YrWBK-L.jpg

Coralnutz was right - trying to boost the areas of the dark car brings out the noise inherent in those dark areas. If you shoot a sequence of exposures at a low ISO and properly expose the various areas of the image across that exposure sequence, then some of your noise issues will be addressed. I have found exposure blending to be quite successful with a series of exposures where noise is a necessary evil.

"Clipping", whether highlight or shadow, means exposing your image so that information is lost. In the case of highlights, it means that the exposure was set so that the highlights (like specular glints of sun off of chrome or glass, etc., or bright spots of light like the white wall of a house) are too bright and get "blown out" - this is all just a shorthand way of saying that your image lost information in the brightest areas because the exposure let too much light onto the sensor and those areas got "clipped" or turned to pure white - even though in reality they may not have even been white. To "expose to the right" refers to the right hand side of the histogram, or the plot that you often see on the back of a camera display or in photoshop books that displays the levels of luminance from 0 to 255. Think of a histogram this way (for a simple 8 bit image): imagine you have 256 buckets labeled "0" (on the left side of the plot) to "255" (on the right side of the plot). For each pixel in your image you measure its luminance (brightness) and then drop that pixel in the bucket that corresponds to its luminance value. You go through your whole image, stacking the various luminance values of each pixel into the appropriate bucket. The brightest values go in the "255" bucket. Anything too bright to be displayed (i.e., an overexposed area in your image) will also go in the "255" bucket. Thus, clipped highlights will appear all the way on the right hand edge of the histogram, usually as a big spike. If you have the "blinkie" feature enabled on your camera's display, these blown out areas will "blink" on the preview of your image when viewed on your camera, to tell you those areas have been overexposed.

The nature of the camera's sensor and the way in which data are treated is such that the right half of the histogram, where the midtones and highlights live, contain more levels of luminance than the left half of the histogram (dark midtones and shadows). When you "expose to the right" what that is saying is to expose your scene so that the data in your histogram live more in the right hand side of the histogram, where more levels of luminance are captured. You do this to get more data to work with, but you still don;t want to blow out, or clip, your highlights. So you expose to the right without overexposing. Then you can post-process your image data and redistribute the tones to your liking.

With a sequence of exposures combined into a true HDR, you are taking the non-blown-out luminance data from several serial exposures and combining it into a single set of data. In this way, the light-capturing ability of your sensor is effectively extended over a greater dynamic range than it could actually capture in a single shot. This is why "HDR Images" made from a single RAW capture are not really HDR - that is, they are not images produced from an HDR dataset. They are really just tonemapped images from a single exposure that have what people seem to refer to as the "HDR" look because of the effect you can get from the tonemapping operators that come with HDR apps. Combining several TIFFs generated from a single RAW image, via tweaking the exposure with the exposure slider in ACR, is only moving the single image dataset around the histogram, it is not actually creating any more data. Sometimes a tonemapped, HDR-ish image generated from a single RAW is called a "pseudo-HDR" image.

http://kirkt.smugmug.com/photos/439100904_eqfNP-L.jpg

Here is a sequence of images used to make an HDR dataset, and the tonemapped result - it is a conference room with a bank of windows with a brightly lit sky. The image is of a 2.5" chrome ball bearing placed on the table in the room, so you are seeing the reflection of the scene in that ball bearing's surface - a cheap way to capture a 360° view of the room. The tiny images down the left side of the frame are small versions of the sequence of exposures that were shot and combined to assemble the HDR. You can see that in the upper exposures, the sunlit window area is blown out, but the table woodgrain, etc. is exposed properly. In the lower exposures, the sunlit windows are exposed properly, but the rest of the scene is buried in shadow and underexposed (clipped). When combined and tonemapped, you get an image that reveals all of the data, an image that your camera could not effectively capture with a single shot - i.e., an HDR image.

Verbose, but hopefully helpful.

Have fun!

Kirk

[JoYork]

Excellent reply, Kirk - that post should be made into a sticky :)

[golclub]

Hello, Kirk! Many thanks for shedding light on histograms!

Thanks also for working with my image and showing what can be achieved through FDRTools. I am curious how you could get rid of "noise" there!!! I just downloaded FDRTools Basic and have to to learn some ropes.

Kirk, when you said "Combining several TIFFs generated from a single RAW image, via tweaking the exposure with the exposure slider in ACR, is only moving the single image dataset around the histogram, it is not actually creating any more data. Sometimes a tonemapped, HDR-ish image generated from a single RAW is called a "pseudo-HDR" image." does it really mean that if I work with a single RAW, I should not really make several TIFF files with different exposures from the file and then combine them? If not, what the best way of treating one single RAW file?

THANKS A LOT!
Maxat