Single Image HDR An often asked question maybe?

russellsnr

russellsnr

Suspended / Banned
Messages
3,121
Name
Russell
Edit My Images
Yes
Hi, Have posted this on another forum as am trying to get a wider view on the subject.
OK the question "Is an HDR image from three exposures that much better than from one adjusted RAW file"?
I have been looking on-line for information in the difference between 3 x images and the 1 X RAW image adjusted to for example 2 stops over (saved) 2 stops under (saved) and most information I have found is around two years or more old.
With the camera companies now offering sensors with mega pixels at 24 to 30 plus is it really necessary to bracket for HDR?
I personally use a hand held meter for exposure that I calibrated to the camera that I use (Canon 1DS MK II not a new camera but 16MP) so when I push the button I have an image that is ETTR without any lose of detail. The software tells me that my sensor can include approximately 3 stops over and 2.5 under on the exposures and as most seem to set there cameras for +-2 stops for three bracketed exposures this leads me to ask the question above.
I hope you can understand what I am trying to express here and welcome any comments/thoughts on the single RAW file HDR.
Thanks
Russ
 
When I did HDR to me if was always better to use real bracketed photos as you get less noise, however I cannot help but wonder if the detail is effectly there in the file to get the files for a triple backed why are you needing to HDR, I find these days better ability to process a file and the dynamic range of my sensor means I havn't processed a HDR in getting on for 2 years

I don't see as the sensor pixel count is going to have on the need to do HDR other than possibly increased noise on the more very tightly packed pixels of a large MP sensor
 
What you can achieve with or without HDR bracketing depends on the effect you want, the dynamic range of the sensor, and the dynamic range of the scene. It has nothing to do with pixels, though larger sensors are better as they collect more light/photons that extends dynamic range deeper into the shadows, with less noise.

FWIW, I hardly ever use bracketed HDR these days (Canon 5D2) as I can get what I want from careful expose-to-the-right (ETTR) settings and then lifting the shadows in Lightroom. That usually works well at low ISO, and indeed I sometimes hold things back a bit if it's beginning to look 'too HDR' if you see what I mean.

Also, no exposure meter can give you optimum ETTR settings. You can only do that by checking the histogram/blinkies and making individual judgements according to the subject. ETTR is not a fixed amount of over-exposure - some benign scenes might allow as much as three stops of extra exposure (my experience) while others may not have room for any more at all.
 
HoppyUK said:
What you can achieve with or without HDR bracketing depends on the effect you want, the dynamic range of the sensor, and the dynamic range of the scene. It has nothing to do with pixels, though larger sensors are better as they collect more light/photons that extends dynamic range deeper into the shadows, with less noise.

FWIW, I hardly ever use bracketed HDR these days (Canon 5D2) as I can get what I want from careful expose-to-the-right (ETTR) settings and then lifting the shadows in Lightroom. That usually works well at low ISO, and indeed I sometimes hold things back a bit if it's beginning to look 'too HDR' if you see what I mean.

Also, no exposure meter can give you optimum ETTR settings. You can only do that by checking the histogram/blinkies and making individual judgements according to the subject. ETTR is not a fixed amount of over-exposure - some benign scenes might allow as much as three stops of extra exposure (my experience) while others may not have room for any more at all.
Click to expand...

Hi, Firstly thank you for the reply.
I have to disagree with you on the 'no exposure meter can give you optimum ETTR'
As you probably no the first thing a camera does for exposure is to try and find middle grey 18% it cannot distinguish anything other than what it sees through the lens where as a hand held meter looks at the whole light falling on the scene or a 1deg point of the scene so gives a more accurate reading of light.
The histogram is only a Jpeg image and the blinkies are very conservative as decided on buy the camera manufacturer IMO.
Also the meter can give you that ETTR exposure to within a hairs bredth depending on how you set up the meter to the camera.
The software that comes with the meter allows you to adjust the meter after reading from an image taken with the camera, I can take an image knowing that the highlights will be as far to the right without reaching that 255 point of no detail.
I understand what you are saying but why use individual judgment? when you take a meter reading put the results into the camera and not even have to look at the back of the camera and no that the image will be perfectly exposed when you put it on the computer?.
As for the HDR what I am looking for is to get that realistic image, the one I saw before I pushed the button, my big head ache is the weather where I live, bright blue skies half the year with not a lot of cloud, it causes a lot of frustration sometimes as the bright to the dark is a very large gap sometimes.
Again thank you for the reply I do not just read it and forget I try and take all on-board to help improve my photography.
Russ
 
russellsnr said:
Hi, Firstly thank you for the reply.
I have to disagree with you on the 'no exposure meter can give you optimum ETTR'
As you probably no the first thing a camera does for exposure is to try and find middle grey 18% it cannot distinguish anything other than what it sees through the lens where as a hand held meter looks at the whole light falling on the scene or a 1deg point of the scene so gives a more accurate reading of light.
The histogram is only a Jpeg image and the blinkies are very conservative as decided on buy the camera manufacturer IMO.
Also the meter can give you that ETTR exposure to within a hairs bredth depending on how you set up the meter to the camera.
The software that comes with the meter allows you to adjust the meter after reading from an image taken with the camera, I can take an image knowing that the highlights will be as far to the right without reaching that 255 point of no detail.
I understand what you are saying but why use individual judgment? when you take a meter reading put the results into the camera and not even have to look at the back of the camera and no that the image will be perfectly exposed when you put it on the computer?.
As for the HDR what I am looking for is to get that realistic image, the one I saw before I pushed the button, my big head ache is the weather where I live, bright blue skies half the year with not a lot of cloud, it causes a lot of frustration sometimes as the bright to the dark is a very large gap sometimes.
Again thank you for the reply I do not just read it and forget I try and take all on-board to help improve my photography.
Russ
Click to expand...

A meter reading, incident or reflective, cannot make the subjective/qualitative judgements necessary for optimum ETTR technique. Only the photographer can make those decisions, based on the subject and the dynamic range available with a particular camera, and using the histogram and blinkies for reference. Some knowledge and experience is necessary, and some testing to know where the limits of the equipment lie. ISO also has a significant impact on dynamic range.

If the subject is predominantly dark, with no bright highlights, then ETTR allows you to add tons of exposure that will greatly enhance shadow detail. But if the subject is mainly light, high-key tones, then there may be no opportunity for any extra exposure without blowing it to bits. Exposure meters take no account of different subjects like this.

In practise though, all subjects contain a variety of dark and lighter tones, and even when they're mostly dark like the proverbial a black cat in a coal cellar, there will be some much brighter areas - highlights on shiny fur, eyes, teeth etc. Specula highlights (reflections of the light source) are often very bright indeed compared to the image overall and the key to optimum ETTR technique is deciding which ones are import and which ones to let blow.

For example, in a wedding group there will be specula highlights all over the place - white collars and cuffs, shiny white paintwork in the background, the bride's dress, and shiny foreheads, cheeks and noses. And these highlights will not be the same brightness either, far from it, but some are clearly much more important than others. Let the bride's dress blow at your peril, though she'll probably be wearing make-up that will reduce reflections off her face; never mind the groom as he's not so much of a worry either way; but you don't want to play too safe because the mother of the bride is wearing a big hat and her face is in shadow. If you want to play the ETTR game, and it's often not for the faint hearted, then use the best tools you can - histogram and blinkies plus knowledge and good judgement, but not an exposure meter.
 
Last edited:
russellsnr said:
As for the HDR what I am looking for is to get that realistic image, the one I saw before I pushed the button,
Click to expand...

If it's a realistic image you want, then yes, taking multiple exposures will be beneficial. Even of the dynamic range of the sensor covers the dynamic range of the scene, you're still taking what is essentially shadow detail at the edge of that range, and moving it further up the histogram. It adds noise. If you are even remotely concerned with quality, you take a set of exposures that covers the dynamic range of the scene, preferably by spot metering the darkest and lightest to establish that exposure range (so essentially placing the darkest and lightest on what Adams would have called zone v). With multiple exposures there's no real excuse to be relying on detail hidden at the extreme left or right of the histogram to add noise or crushed highlights to your image. A single exposure image will always contain more noise where the shadow detail has been increased, and will always have some area of highlight that suffers.

In reality though... trying to bring absolutely everything in range can just look awful. I mean.. if I was taking a shot, at evening, or an outdoor café scene, then I would want the actual lamps that are lighting the interior of the café to be blown out. I wouldn't want them so perfectly exposed I could read the make and wattage of the lamp would I? That would look ridiculous and it's an impossibility in real life. Sometimes, you want something to blow out. Shots where everything is captured perfectly just look stupid sometimes.
 
russellsnr said:
With the camera companies now offering sensors with mega pixels at 24 to 30 plus is it really necessary to bracket for HDR?
Click to expand...

Actually, it's the other way round - increasing the pixel count makes it harder to get a high dynamic range from the sensor. For a given sensor size and a given lens, if you increase the number of pixels, you decrease the physical area of each pixel and, hence, the number of photons landing on each pixel. You therefore need more gain in the pixel's amplifier, which increases the noise reducing the Dynamic Range. If you increase the sensor size, but keep the pixel count the same, same result.

This is why you can get 14 stops plus from a full frame camera and only 10 from a compact - same number of megapixels but larger pixel sites getting more photons.

If you have a 14 stop camera and bracket +/-2, then you get an 18 stop image. Whether you need this or not depends on the dynamic range of the scene.

To store and process the image, you need enough levels (and therefore bits per pixel) to ensure that you don't get any visible contours.

Finally, you need to display it. At the moment, we tonemap the HDR image back to an SDR image capable of being displayed on a monitor or paper.

In the near future, expect monitors with much higher peak brightnesses and more bits per pixel to allow the proper display of HDR images.
 
st599 said:
Actually, it's the other way round - increasing the pixel count makes it harder to get a high dynamic range from the sensor. For a given sensor size and a given lens, if you increase the number of pixels, you decrease the physical area of each pixel and, hence, the number of photons landing on each pixel. You therefore need more gain in the pixel's amplifier, which increases the noise reducing the Dynamic Range. If you increase the sensor size, but keep the pixel count the same, same result.
Click to expand...

In theory, raising the number of pixels 'should' reduce the dynamic range, but according to DxO the 16mp D4S has a dynamic range of 13.3 Evs, the 24mp D610 has 14.4 Evs and the D800 also has 14.4 Evs. As the pixel counts have increased they have reduced the gaps between the photosites and played about with the micro lenses to offset some of the need to reduce the size of the photosites. So I have read anyway. ;) What they seem to get with the larger photosite is increased high ISO performance. There will probably become a point, if we are not there already, where higher pixel counts will detrimentally affect dynamic range, and high ISO performance.

My camera is about 5-6 years old and a DX sensor, so I do not have the dynamic range or noise performance of modern sensors. When I have done HDR from one processed RAW file it was because I didn't plan a HDR when I took the image, or there was something moving where a HDR from bracketed images wouldn't work. One image HDR does increase the noise a lot, so normally using bracketed image should give a better end result.

I have found that processing with the later versions of Camera Raw or Lightroom negate the need to try HDR from one image most of the time, and it isn't something I have done for a year or two.
 
It can and has to be done in certain circumstances (where the subject is moving).

If I can't bracket, I will expose to the left and pull back in PP creating multiple images. I believe this shot was taken on a D5000 at ISO200 - a camera with a couple more stops DR will give a much better result but just to give an idea...

The reason I wouldn't personally ETTR is because it's easier to pull back detail from shadows 4/5 stops but if you have blown the hightlights by 2 stops, it will look horrendous when trying to make the final image.

Egypt sunset over the mountains by PhilYoung1, on Flickr
 
I now most HDR processing software have algorithms to try and deal with movement, but it depends on how much movement, and what is moving. :) What it can and can not deal with will only come from experience with the software.

As long as you don't overexpose, or at least not too much, I'd always ETTR. One of a set of bracketed images could be the perfect ETTR RAW image too.

This article from a Adobe shows how the tones are distributed in a RAW file, and in which area the most data is, and it's not in the shadows. ;) And the reason why people use ETTR to try and get the most out of an exposure.

If it works for you though, fine. :)
 
redhed17 said:
In theory, raising the number of pixels 'should' reduce the dynamic range, but according to DxO the 16mp D4S has a dynamic range of 13.3 Evs, the 24mp D610 has 14.4 Evs and the D800 also has 14.4 Evs. As the pixel counts have increased they have reduced the gaps between the photosites and played about with the micro lenses to offset some of the need to reduce the size of the photosites. So I have read anyway. ;) What they seem to get with the larger photosite is increased high ISO performance. There will probably become a point, if we are not there already, where higher pixel counts will detrimentally affect dynamic range, and high ISO performance.
Click to expand...

Looking at the numbers, you can either maximise the gain of the pixel amplifier ("ISO") or its low noise performance (which affects the dynamic range). Looking at 2 cameras from 5 years apart on there - there seems to be a trend towards higher ISOs and slight increases in dynamic range.

When the displays that are being shown at the likes of IFA and CES - high brightness with enough bits per pixel and contrast ratios to show HDR images without tone mapping/contouring/.... - become available, cameras which can exploit them will soon follow.
 
redhed17 said:
I now most HDR processing software have algorithms to try and deal with movement, but it depends on how much movement, and what is moving. :) What it can and can not deal with will only come from experience with the software.

As long as you don't overexpose, or at least not too much, I'd always ETTR. One of a set of bracketed images could be the perfect ETTR RAW image too.

This article from a Adobe shows how the tones are distributed in a RAW file, and in which area the most data is, and it's not in the shadows. ;) And the reason why people use ETTR to try and get the most out of an exposure.

If it works for you though, fine. :)
Click to expand...
Hi, Agree 100% there is more detail info in the highlights of an image that is one of the reasons as I said in the OP that I like to use a hand held meter.
Thanks for the replies.
Russ
 
Right, here we are...

Exposed to the left and to the right, loaded the RAW's into CR and made different exposures, loaded into photomatix and blended using the same preset.

Photo 1 ETTL:
10342474_10154240790200305_2215507782332200686_n.jpg


Photo 1 ETTR:
10372034_10154240790935305_3617476630802671744_n.jpg


Photo 2 ETTL:
10366271_10154240791205305_5273142958897630344_n.jpg


Photo 2 ETTR:
10377995_10154240791640305_9099325886198165232_n.jpg


Not much difference in #2 where the dynamic range isn't as huge but #1 is definitely better under exposed then pulled back. This is the technique I would use for reasons said above but feel free to do your own tests for your own piece of mind!

*The Jesus writing isn't mine!!!!!
 
Phil Young said:
Not much difference in #2 where the dynamic range isn't as huge but #1 is definitely better under exposed then pulled back.
Click to expand...


Highlights may be..... but I bet shadows aren't, and can I suggest in that shot, the shadows are more important?


This issue, and ones like it, can be sorted out with just taking two exposures and blending them. Why resort to tone mapping at all? Tone mapping just makes stuff look weird.
 
Last edited:
Pookeyhead said:
This issue, and ones like it, can be sorted out with just taking two exposures and blending them. Why resort to tone mapping at all? Tone mapping just makes stuff look weird.
Click to expand...
I could have processed it further to make it more authentic but that's not the purpose of the test.

As for what's important, in a HDR shot, everything is.... to me at least.
 
Last edited:
Hi, OK I took this today and used the Magic Lantern file that you can put on a CF or SD card to allow a Canon camera to take from the 3 to 9 exposures, I set mine to 7 at 1 stop apart, took the image from Lightroom into HDR software to blend then back to Lightroom for W/B and set white and black levels.
What I need to no is does it look realistic or should I have done something else to push it a little further?
Advice welcome and thank you for the above replies.
Russ
 
It looks okay Russ, a bit flat perhaps. Hard to say without some kind of comparison shot, but for my kind of realism (that may be different to yours) I'd expect to see some blown highlights in the curtains, and some pure blacks, in the near plant perhaps. HDR has also done something strange with the shadows around the far plant, and created some fake lighter bits with an odd hue. And the floor has changed colour.

I would just say that 'realism' is subjective and simply capturing max dynamic range is no guarantee. Cameras don't record things the same way we see and perceive them. As Pooks says, leaving it to HDR technique and software often ends up with something that looks far from real, especially when you try too hard and take it too far.
 
Last edited:
You must log in or register to reply here.
Back
Top