Active D-Lighting no more than highlight protection and Shadow?

[lorriman]

I'm the sort of chap who wants to do the minimum pp, so this kind of function would be jolly useful to me.

[ellis_vener_photography]

MS keil is correct in his description of how it works.

[sven_felsby]

Yes, Greg, I agree 103% - so much more fun to shoot in the field than fiddling at the computer! I also want ADL, but must wait until I drop my beloved D200 in front of a steamroller.....

[allardk]

Ellis, you have me confused. MS Keil's description does not sound to me like a <i>'step that is done on the chip

on a photo site by photo site basis at the time the light is being recorded by the individual light sensitive

cells in the CMOS'</i>. It sounds like a step that is done in the in-camera raw conversion. Not one that is as

easily reproduced in pp as a curve adjustment, but to me it sounds this happens to already recorded data.

<p> I'm assuming the steps in image production are: exposure->readout->amplification->A/D conversion->Image

processing->writing files.

<p>You seem to be well-informed on this. To your information does ADL involve operations on individual pixels (or

small regions) before A/D conversion?

[ellis_vener_photography]

"Ellis, you have me confused. MS Keil's description does not sound to me like a 'step that is done on the chip on a photo site by photo site basis at the time the light is being recorded by the individual light sensitive cells in the CMOS'. It sounds like a step that is done in the in-camera raw conversion."

 

Okay you are a victim of my bad writing. but it is done on the CMOS on the individual photosite level as the signal travels from the photosite to the EXPEED processor.

[jakob_norstedt_moberg]

S Y, I made a similar experiment with my D3.

With exposure set to M and similar exposure time on both, and quality set to NEF+JPG I get the ADL effect on the JPG:s. But the RAW files are absolutely identical when viewed in Adobe Camera Raw 4.5. I see this as an evidence that the ADL is an effect created not in the individual CMOS photosites but in a later process step.

 

Ellis, Please give a trustable source for the statement of yours that it is done on the individual photosites. How do you explain the experiments of S Y and myself?

[elliot1]

"But the RAW files are absolutely identical when viewed in Adobe Camera Raw 4.5"

 

This is normal. You need to use NX or NX2 to see ADL and other Nikon in-camera 'enhancements'. Which is odd if the modifications are made on the sensor. Hence my original comment that I doubt the changes are actually made on the sensor.

[jakob_norstedt_moberg]

Elliot, exactly my point.

 

Jakob

[philippartridge]

From the material presented thus far in this thread, and I confess I know little about the ADL feature, it may be targeted at a straight compression of algorithmically determined 'out of bounds' values for shadows and highlights to bring each pixel value proportionately into the sensor's DR - for in-camera jpg production, which need all the help they can get.

 

If so, one could presume the effect was calculated after the effects of tone curve, colour space and sharpening settings were factored in, and that the result is simply the scene with attenuated or increased luminance levels, but with all other settings intact; and with the calculation performed at the photosite level.

 

Questions it raises include: can one adjust the threshold (forgive me if this is known)?, and how, and how well, does it deal with specular highlights, which of course can be an image feature?

 

The photosite level calculation may not be so fancy, as a curve or blended mode adjustment does not affect each pixel uniformly, but does so commmensurate with its value..

 

As a RAW shooter, I would like instead an accurate measure of the extent of 'apparent over-exposure' (AOE) I can get away with shown on the LCD histogram's jpg preview - it (the LCD) could be made colour-aware, could simulate the amount of AOE it displays to reflect the dynamic range of sensor. One can get close using experience but if time is tight, I often need to move on quickly than laboriously re-shoot and retain several exposure versions of the one image.

 

Final comment: Shadow/Highlight is a pretty rough tool for restoring credible shadow detail, IMO; much better is the use of shadow luminosity masks in conjunction with the screen blend mode, applied judiciously; which delivers well with low levels of artifacts.

 

Thanks for the Nikon information, Ellis.

[jakob_norstedt_moberg]

Philip, Ellis et al,

 

If the NEF pixels aren't changed by turning on ADL how could it be a calculation at the photosite level? You do not seriously mean that there are two different flows of information from each photosite; one for the NEF and one for the JPG? That is really a way of complicating things.

 

Please show me an evidence of this being the case, not just speculations from rumors.

[philippartridge]

I will try to clarify my point, Jakob. I was only speculating about the method used by ADL, wondering aloud how sophisticated it was. At the end of the day, it is just another variable, an attribute to deal with in the mix in post-processing for a NEF (RAW) file, is it not?

 

I doubt any processing physically occurs at each photosite, but that each value may be dealt with in a matrix computation on the chip, and that this could be quite smart and sophisticated or it may be more in the manner of a creation of the public relations department, i.e., no big deal. The way Nikon have been adding characteristics to image processing, it must be leading to analysis paralysis for many!

 

I for one dislike having to devote days to camera manuals, but, shrugging shoulders, it is necessary. Features do vary enormously in usefulness though. I remember a Microsoft trainer telling an audience that '50% of the features people asked for in Word are already in the software'. The art of software design (all those shooting banks and the menu contents, for example) is visibility and prioritisation of features. Nikon are apparently ahead of Canon in this respect...

[jakob_norstedt_moberg]

Philip, I gree with you. /Jakob

[frans_waterlander]

Back from being gone for a couple of weeks. After reading all that has been said here, and thanks to everyone responding, it appears to me that NOBODY has really explained what ADL is really doing. It still looks to me that ADL is nothing more than a reduction of exposure before the shot is taken to reduce the chance of blowing out highlights and a tonality curve tweak after the shot is taken to salvage some of the shadow information that was compromised by the reduction in exposure. And don't get me started on this photo site by photo site manipulation, because that sounds like a bunch of hogwash (with my apologies to the pig family).

 

So, it looks like Nikon has done an excellent snowjob in that they have introduced a questionable "feature" and have many believe it's better than sliced bread.

 

I'm still looking for an honest and accurate description of exactly what ADL does from an authoritative source.

[Brad_]

>>> And don't get me started on this photo site by photo site manipulation, because that sounds like a bunch of hogwash

(with my apologies to the pig family).

 

Well, since your mind is closed, I'll pass on explaining again. You can search on previous threads if you care to.

[frans_waterlander]

"I'll pass on explaining again"

 

Brad, maybe I'm missing something here but try as I may, I don't see any previous response from you.

[daniel_kalchev]

<p>After much debate it seems very likely that Nikon records a dump of the 1005 pixel RGB sensor in the NEF file. This information can be used later to process the different 'regions' with different tonal curves much like the Apical patent (and their software and hardware) describes.<br>

The availability of this metadata explains two things:<br>

1. Only Nikon software understands/processes Active D-Lighting.<br>

2. The RAW data contained in the NEF file is similar, with or without Active D-Lighting.<br>

I too was under the impression that something is done at the sensor level, because of a document (now unavailable) from SONY which was discussed when they announced the A700 and DRO.<br>

Doing the same processing at the sensor level is perfectly possible but brings out the issue that your RAW data is no longer 'raw'. Other software will have hard time applying sensible tonal curves. Especially if it does not have access to the 'map' from the metering sensor.<br>

Whether applying up to 1005 different tonal curves to different regions of the image is sufficient is another story - it is likely that Nikon will increase the density of their RGB sensor to much larger number of pixels in a next generation cameras.</p>

[frans_waterlander]

<p><strong><em>"Doing the same processing at the sensor level is perfectly possible"</em></strong><br>

As I have argued in the past, this is not easily done and very likely not the case.<br>

<strong><em>"it is likely that Nikon will increase the density of their RGB sensor to much larger number of pixels in a next generation cameras."</em></strong><br>

What do you base that assertion on?</p>

[daniel_neo]

<p>It is a feature, like auto gear in a car. For those who just like to just enjoy cruising along without the hassle of gear shifting (in this case Post processing your images), then ADL is a heaven-sent (or Nikon-sent) convenience, just go for the Auto. For those who is particular about how their car behave and who want to have 100% control of their cars, feel free to go for the manual cars, ie, set ADL off and doing all the adjustments yourselves.<br />With ADL you gain convenience, but loses some amount of control, which for most people is no biggy. Small price to pay for the convenience.<br />I'll set it on, but for important images that I foresee doing certain amount of manipulation, I'll capture RAW+JPG. If it turns out fine, I'll just use the JPG. If I need to post process it, then I'll have the RAW file to play with. Well, the RAW file me may be up to 2/3 EV below the exposure I consider optimum, but hey, 2/3 EV (ADL High) ain't that big a deal when working with RAW files.<br />That's just my 2 cts. Thanks.</p>

<p> </p>

[mike_morris3]

<p>I have read through this discussion carefully, along with other discussions of this topic on other web sites. I have reviewed the patent summary mentioned by MS Keil above. It appears to me, from that patent discussion, that Active D-Lighting is an algorithm that is applied on a region or area basis. I do not consider myself knowledgeable enough to make any statement about <em><strong>where</strong> </em> the algorithm is applied (i.e., on a pixel level or elsewhere).<br>

What I have not been able to determine from reading this and other forums, is whether the algorithm is included or not in the RAW image data. It is possible to conclude, as Daniel Kalchev has above, that Nikon RAW files are "...no longer 'RAW'"--if Active D-Lighting is enabled.<br>

I would certainly like to locate more information on this specific aspect of Active D-Lighting. Can anyone help?<br>

I have seen a number of histograms that show subtle differences, but, unless I am wrong, that does not answer the question about RAW image content.<br>

If anyone is aware of any technical documentation of this topic (as it applies to Nikon's implementation), i would like to know about it--I have a background in Electrical Engineering and signal processing, so engineering documents are acceptable.<br>

BTW, for what it is worth, the range of evaluations of this feature displayed in this forum--from "Never Use It" to "Use It Cautiously" to "Always Use It" is repeated in other forums.</p>

<p> </p>

[frans_waterlander]

<p>As far as I can tell ADL is applied to the RAW file if it is activated. However, that doesn't mean more than my best guess as apparently Nikon has gone through great pains to promote this feature without telling anybody what exactly is going on.</p>

[pete_s.]

<p>Frans,<br>

It may not be too difficult to find out if it is applied to the "raw" image or not.<br>

Just shoot two images from a tripod in manual mode, one with heavy ADL and one without. Check them out in the free rawnalyze software. If the histograms are vastly different the ADL affects the raw values.</p>

<p> </p>