Nikon dSLRs builtin HDR

<p>Hi, I read that the newer bodies have HDR built into the camera body. </p>

<p>I've not liked HDR s/w personally due to ghosting with low light photography due to long shutter speeds.</p>

<p>Question:<br>

How is the in camera HDR like? Do you get ghosting?</p>

<p>Thanks.</p>

<p>No, SLRs have had the exposure bracketing feature such that you end up with several images with different exposures. If you choose to, you can then merge those images together using HDR (high dynamic range) during post processing.</p>

<p>Some cameras have this feature built-in, but they are not Nikon DSLR's. They will still have to be done this way, by taking several shots at different exposures (shutter speeds), and merge them together using software algorithms. Ghosting does appear if camera/subject movement is involved. The degree of customizing the tone curve will be very limited when done in-camera versus in post processing.</p>

<p>Newer Nikon dSLRs have "active d-lighting", which adjusts the tone curve in cases where the dynamic range of the scene exceeds what would normally be output; it can recover some highlight and shadow detail in a similar way to HDR imaging. However, it only works with one capture, and is therefore limited to the dynamic range of the sensor (which is typically several stops beyond the range of a JPEG); you can do the same kind of thing with raw processing.<br>

<br>

To get more dynamic range than the sensor can capture in one go, you do need multiple exposures, which some other cameras can do automatically and Nikons can only do by bracketing and manual post-processing. I have suggested in the past to Nikon that it might be nice to have a mode wherein the normal capture detects that it's run out of highlight range, and automatically fire the shutter a second time for a shorter exposure so that the highlight information is available if a user wants it; so far you still have to select bracketing manually.</p>

<p>Multiple exposure is the way to improve your pictures, but with limitations.<br>

Nikon does not call it HDR, but you can take from 2 to 10 exposures that will be "collapsed" into one picture, automatically. Possibly providing better colors and dynamic range. All done by the in-camera firmware, and no post-processing is needed.</p>

<p>See if it works, or not, for your application.</p>

<p>Read page 184 in the D300S Manual.<br>

This was already discussed here, few times in the passed.</p>

<p>Frank - I didn't think you could combine multiple exposures automatically with exposure bracketing. If you can, it's not clear from the D700 manual (although the D300s may differ). If there's no way to alter exposure, I can imagine that you'd get better noise handling by (effectively) averaging frames, but I doubt it'd give you much of a tonal advantage. I'm prepared to be wrong - I'll admit that I've ignore the multiple exposures feature, and only rarely used bracketing.</p>

<p>I think what is meant by in-camera HDR is what Nikon calls Active D-Lighting.</p>

<p>Active D-Lighting is a good idea, it works in certain situations. But it underexposes the image and can add noise, which in most cases isn't a good thing. It's better to leave it off and use the D-Lighting feature in Nikon Capture NX2 after the fact, which works pretty well.</p>

<p>I was wishing for in-camera hdr a few years ago. The D-lighting is wonderful, but it's a compensatory function to me, and not an effect. And it's not perfect, because of what looks like noise.</p>

<p>You need to read camera manual carefully, and try to understand what is there...<br>

Look up your camera menus. You need to know your cameras completely. Do not skip on learning.<br>

I am telling you what Nikon's HDR is implemented like, as in D300S camera, and yet you do not believe me.<br>

Yes, D700 has the same Multiple exposure (or call it HDR), implemented.<br>

Read on page 198 in the D700 manual.<br>

The question is when to use it to take maximum advantage of it. Is not for all picture shooting.</p>

<p>I don't own a D300 or D700, but I wonder: Active D-Lighting may add noise in the D300's images, but surely not so much in the D700's images? And post-processing can reduce noise, right?</p>

<p>Frank - apologies if I offended you; I'd not meant to sound disbelieving.<br>

<br>

My concern was that I'd read the relevant section (as you say, page 198) in the D700 manual, and indeed it talks about combining multiple frames into a single image. The way it's described suggests that it's a simple blend of the specified number of frames, and that all the frames are evenly-weighted; I've not found any discussion about how multiple exposures might combine with bracketing, or that there's a way of selectively mapping the contribution from different exposure settings.<br>

<br>

In other words, I agree there's a multiple exposure mode (that combines the exposures in camera), I'm just confused how this can be used to generate HDR images; it doesn't appear to be sufficiently configurable.<br>

<br>

It may be that this does, in fact, produce HDR output - unfortunately I only have access to the manual with me, not my camera, so I can't tell whether it works but the documentation is confusing me.<br>

<br>

I'm not saying that you're wrong - I'm just saying that I can't see how to use this feature for HDR. I will always assume that the error is in my understanding, not yours.</p>

<p>Hi Andrew,</p>

<p>.. no need to apologize, and I do not offend that easily...</p>

<p>just did not like some statement that some unspecified other camera vendors do have HDR but Nikon does not, that was not really true, especially when no specific other camera is named.</p>

<p>In camera HDR is programmed by selecting number of exposures taken, and that happens before they are taken, therefore it is not reasonable to apply other then equal amount of light from each one, since the picture was not taken yet and is not known. Only in post processing, when inspecting each picture for the HDR, operator can choose a different amount of each exposure contribution to the final picture. I do not expect the cameras to provide post processing of HDR, and therefore, this method in very limitted. </p>

<p>Even if the multiple pictures are taken, as opposed to multiple exposures, the HDR has severe limitations to shooting static subjects in perfectly stable condition, over the time period of the total time of all exposures taken.<br>

Nikon's documentation is always hard to read or discover what it really means or how to use it.</p>

<p>Hi Frank. I'm glad no damage was done!<br>

<br>

I assume the OP was talking about something like the multi-shot HDR feature of the <a href="http://www.dpreview.com/reviews/SonyNex5Nex3/page11.asp">Sony NEX series</a>. I don't know how useful that actually is; frankly I always shoot RAW and post-process anyway, so it wouldn't matter much to me even if Nikon did support it in-camera. As far as I know, no Nikon automates this process in the way that the Sony does, except in as much as active d-lighting does it within one frame.</p>

<p>Per DPReview link Sony does:<br /><em>"Sony's 'Auto HDR' feature that takes multiple shots of the same scene at different exposures, then combines them to produce a single image that incorporates a larger range of tones than would be possible from a single exposure"</em><br />This is exactly what Nikon Multiple exposure does.<br />Also Sony:<br /><em>"The breadth of the exposure gap between these three shots can either be chosen automatically by the camera or set manually.</em><br />Nikon only allows automatic breadth selection of the exposure gap, and does not allow manual entry, so Sony seems to have more flexibility here.</p>

<p>However, in Sony:<br /><em>"Sadly Auto HDR isn't available in iAuto mode (it'd be a real benefit to users who just want to capture a scene rather than worrying about how high contrast it is), and it requires a lot of button pressing, particularly if you want to manually select the extent of the effect."</em><br />For Nikon this is automatic mode.</p>

<p>Then for Sony:<br /><em>"the NEXs take three images</em>" - while Nikon takes any number between 2 and 10.</p>

<p>Andrew, I am surpsied with this:<br />"<em>As far as I know, no Nikon automates this process in the way that the Sony does, except in as much as active d-lighting does it within one frame."</em><br /><em></em><br />Definitelly Nikon does it differently than Sony, as explained. <br /><em>...</em> and the<em> "As far as I know"...</em> no comment<em>.</em><br /><em></em></p>

<p>Hi Frank. I'm prepared to accept that I don't know what I'm talking about, and I need to get my camera and try this (which, unfortunately, I keep forgetting to do when I'm at home).<br>

<br>

My impression from the DPReview discussion (and the results) and from reading the multiple exposures section of the D700 manual is that:<br>

<br>

The Nikon multiple exposure mode takes a specified number of exposures, then combines them either by just adding the resulting (RAW) pixel values (auto-gain off) or by scaling the pixel values down by the number of exposures taken and then adding the results (auto-gain on). I get this from p.199 of the D700 manual - it could be a dubious description. Either way, the contribution to each pixel is an evenly-weighted sum of contributions from all the input images. Assuming that the more-exposed regions clamp you do get some dynamic range extension if each image has a different exposure (it's not clear to me whether this can be automated - whether bracketing counts towards the multiple exposure count - or whether you'd have to bracket manually), but there are horrible angles in the tone curve: it turns the tone mapping into piecewise continuous segments.<br>

<br>

The Sony HDR mode explicitly takes three exposures with different levels of illumination, then selects samples from each image in a way that keeps all values within range. Whether this is done by explicitly tweaking the tone curve or (better) by applying a very low-pass luma filter over the image and using that to weight the pixel contributions, I couldn't say, but I wouldn't describe it as an HDR mode if it didn't do something like this - nor would I expect people to be happy with the outcome. It's this variably-weighted contribution that I don't believe the Nikons do, based only on the manual, but I'm prepared to be wrong.<br>

<br>

When you say the Nikons have an automatic exposure gap between shots, I'm even more confused - it is, at least, selectable (although not by much) when bracketing.<br>

<br>

I'll try to give this a go with my camera this evening. Forgive me for flogging a dead horse in the meantime - I'm a computer graphics software engineer, so even if I can get the camera to produce an apparent result, I'm still interested in the theory of what it's doing. Thanks for your patience with me in the meantime.</p>

<p>I suspect at this stage this is only of interest to anyone searching the archive, but...<br>

<br>

I had a play with my D700 this morning. It appears to be impossible to select multiple exposures and bracketing simultaneously - selecting one disables the other. This is presumably why the combination isn't documented in the manual.<br>

<br>

That's not to say that there's no intelligence in the camera to detect when the exposure has been changed during a multi-shot sequence and do appropriate things when blending frames, but since - in the absence of explicit bracketing - I can't imagine how it could tell this scenario from an explicit multiple shot special effect, it seems unlikely to me. As I suggested above, if you manually change the exposure between shots in a multi-exposure sequence you do get a compressed dynamic range due to the clamping of the pixel values from the exposures that are brighter, but it'll be a poor approximation to a "proper HDR" image combined on a computer and tweaked manually. Not that I can vouch for how well the automated scheme in the Sonys work, of course.<br>

<br>

For anyone curious about why I claim there is some HDR effect, here's my understanding. Let's simplify and say we're taking four exposures with an 8-bit sensor and each exposure is - manually - one stop less exposed than the previous. We want to output an 8-bit (brightness) image, as in a JPEG, and - because we're trying to do HDR - we'll say our scene contains elements that would normally map to pixels with values 0..2047 (eight times as bright as we'd normally be able to express).<br>

<br>

For dim pixels that aren't over-exposed in any of these exposures, the pixel value from the most-exposed frame will be some value x in the range 0..255. This is 0..255 in frame 1, 0..127 in frame 2, 0..63 in frame 3 and 0..31 in frame four, for a total of (x + (x/2) + (x/4) + (x/8)), and the result is divided by 4 if we have exposure compensation turned on to average the frames. Note that, for dim pixels, the brightest exposure makes the greatest contribution.<br>

<br>

If the image is bright enough to saturate the first exposure to 255, it will still have a contribution from the later, dimmer, exposures, but the first exposure will contribute a fixed amount. The same is true of successively brighter scene elements as dimmer and dimmer exposures still saturate. Therefore, for a scene element whose brightness would be "x" in the basic exposure, we get the following output pixel values:<br>

<br>

0≤x<256: output (x + (x/2) + (x/4) + (x/8))/4 = (1.875 x)/4 = 0.46875 x, in the range 0..120.<br>

256≤x<512: output (255 + (x/2) + (x/4) + (x/8))/4 = 63.75 + (0.875 x)/4 = 63.75 + .21875 x, in the range 120..176.<br>

512≤x<1024: output (510 + (x/4) + (x/8))/4 = 127.5 + (0.375 x)/4 = 127.5 + .09375 x, in the range 176..223.<br>

1024≤x<2048: output (765 + (x/8))/4 = 191.25 + x/32 or 191.25 + .03125 x, in the range 223..255.<br>

<br>

In other words, the scale factor is piecewise-linear. The more exposures in the final image, the better approximation to a smooth curve you get - but the "angles" will show up as bands in continuous tones in the result. This may or may not be better than nothing; it's certainly not as good as doing this in post-processing. The actual range of the sensor is more than 8 bits and if the image combination is not performed in linear space then the situation is more complicated - but I believe my concern about the output not being C1 continuous still holds in the more complex case.<br>

<br>

I'll be moderately impressed if anyone's still reading at this point, but I'll be interested if my above assumption about how this works is incorrect.<br>

<br>

I hope that helps someone, if only in providing evidence of my insanity.</p>

<p>I generate compressed images manually to approximate "what the eye might see" as automated processes will make decisions which might not "fit" aesthetically. The method has been in common use with astrophotography for decades. For maximum flexibility, one can further split a scene into its LRGB constituents (using filters, at the time of shooting) and apply the technique to each of the acquired layers.<br>

Here's a rudimentary example (see comment section); not very well made but illustrated the technique:<br>

<a href="../photo/4839363">[Link]</a></p>

<p>Many more examples of LRGB + HDR in astrophotography whose technique can equally apply to terrestrial photography:<br>

<a href="http://www.google.ca/images?hl=en&q=LRGB%20M31&um=1&ie=UTF-8&source=og&sa=N&tab=wi&biw=1920&bih=1011">[Link]</a></p>

<p>Michael - very nice.<br>

<br>

To be fair to those cameras that try to automate this, using a low-pass filter to weight the intensity isn't far off what the eye actually does - it's how it manages to express dim and light areas simultaneously. Of course, because you can move your eyes around, you don't see visual artifacts as obtrusively as in a static image. Whether Sony use this technique (the equivalent of dodging and burning) or just flatten the whole dynamic range of the image (low contrast paper, in film terminology), I don't know.<br>

<br>

However, just because it's "what the eye might see" doesn't mean that the automated technique will produce something as nice as the photographer can, working on a computer. Ansel Adams is known for the prints (after a lot of dodging and burning), not for the original exposures. Manually-adjusted HDR is a time-honoured part of photography.</p>

<p>Dear all,<br /> Some comments on the posts:<br /> 1) Current evidence suggests that the retina has at least three mechanisms for adaptation to a scene's luminance range (not <em>brightness</em>, brightness means <em>perceived</em> luminance). First one at photoreceptor level ("global tone curve adjustment"), another at the network level, and a third one (that functions similar to<a href="../nikon-camera-forum/00R3bF?start=20"> D-Lighting</a>) at the horizontal-bipolar cell level with feedback to photoreceptors. Inner retinal adaptation is thought to reduce redundancy ("edge enhancement").<br /> 2) Lowpass filtering for adjusting luminance levels leads to well-known artifacts (see <a href="http://dragon.larc.nasa.gov/retinex/consumer/consumer.html">RETINEX</a>). State of the art algorithms make use of, for example, anisotropic diffusion or erosion/dilation operators (morphological ops).<br /> 3) D-lighting does <em>not</em> adjust the tone curve globally (which would just be tantamount to contrast reduction). It uses morphological image processing operations to define regions, and an individual tone curve is computed for each region.<br /> 4) Have fun.</p>

<p>MS - thanks for the corrections - and interesting RETINEX link. I'd heard of the more advanced techniques being used in computer vision applications, but I didn't know D-lighting was that clever; my apologies to any Nikon engineers whose efforts I've underestimated. I have to assume that the Sony HDR automation is similarly advanced in its tone mapping, and my talk of simplistic techniques dates my image processing expertise to the 1990s. I'm sure there are still advantages, if only in artistic vision, in manual HDR conversion, but this does suggest that the automated system may do rather better than I'd been assuming.<br>

<br>

I've just found your reference, in the <a href="http://www.photo.net/nikon-camera-forum/00R3bF">thread</a> from 2008, to the patent that discusses how ADL works. Belatedly, thank you for posting that - I was about to ask whether there was any more information! :-)</p>

<p>Dear Andrew,</p>

<p>Glad I could help!</p>