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

Based on what I've read here and in other places it seems to me that the much hyped Nikon Active D-Lighting is

nothing more than some mild highlight protection and shadow lightening not unlike the shadow function of

Shadow/Highlight. Now don't get me wrong, I'm an avid Nikon fan and use Nikon cameras and lenses exclusively, but

it looks like one can achieve similar and more than likely better results by watching the highlights when taking the

shot (something that should always be done anyway) and curve manupilation or the application of the Shadow

function. Yes, it takes more attention to what you are doing in the field and in the digital darkroom, but it allows one

to apply an almost infinite range of adjustments rather than have the camera apply some limited predetermined set of

exposure adjustments and shadow tweaks.

 

Some folks have surmized that ADL would somehow tweak individual sensor photo sites depending on the amount of

light at that particular location to modify sensitivity and thus tweak the tonality curve or linearity of the sensor. Sorry

to say, but each photosite on a sensor has about the same sensitivity and cannot be tweaked individually.

 

I'd like to hear your opinion.

My understanding is much like yours. Much the same can be achieved by reduction in exposure and then lifting of the shadows by means of the curves tool.

 

ADL must be especially useful for the rapid production of JPEGs in camera.

If you notice, many Nikon users keep the D-Lighting option off in their cameras and they only apply it on NX2 when needed

it. I think you are right, a good photographer should be able to adjust his shots by means of histogram but the sad truth is

that not all of us are good photographers so for some people it is very useful. Just today, I was trying to get a shot of the

roof of a temple against an overcast sky. No matter how many times I tried to adjust the shot I couldn't get it right and even

though I haven't seen it on my computer yet I know I will need D-Lighting to get a nice print out of it.

"Some folks have surmized that ADL would somehow tweak individual sensor photo sites depending on the amount of light at that particular location to modify sensitivity and thus tweak the tonality curve or linearity of the sensor. Sorry to say, but each photosite on a sensor has about the same sensitivity and cannot be tweaked individually. "

 

Frans,

 

According to Nikon's engineers, that is exactly how it works on a per photo site basisand it is done on the CMOS imaging chip at the time of exposure. Obviously there are limitations - the sensitivity range of the chip and the programming in Nikon's EXPEED processor to keep the results from looking really weird.

 

Rene'

 

Active D-Lighting is diferent from older forms of D-Lighting on older cameras or in post shoot processing.

 

Neither of you have to believe me or them but those are the facts.

 

"I think you are right, a good photographer should be able to adjust his shots by means of histogram"

 

Which histogram? The one on the camera is based on the cameras JPEG settings including compression, color space assignment and any other processing parameters you've set. This is true whether you shoot in camera processed JPEGS or raw for later processing -- so if you shoot raw it simply woon't show you what the ful ldynamic range in the photograph you just shot. The one in your raw processing program or digital darkroom program of choice.

 

If you are intent on getting the maximum flexibility out of you camera and processing hardware, you shoot raw and process the raw out as a 16 bit per channel, large color space (Pro Photo or Holmes D-Cam) TIFF or PSD. And there isn't a digital processing program that displays a 16 bit per channel histogram.

 

Technically I am a very competent photographer and user of digital darkroom software. My attitude is that I'll take advantage of every edge I can get to make my finished photo graphs look the way I want it to.

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

 

Pretty much. And IMO it works great! I leave it on all the time.

Ellis,

 

"According to Nikon's engineers, that is exactly how it works on a per photo site basis and it is done on the

CMOS imaging chip at the time of exposure."

 

Do you have the reference for this claim?

 

I find it hard to believe. I've read the description of the Sony IMX 021 that's used in the D300 and there

doesn't seem to be any reference to this capability. What's more ADL is available on the D60 and that camera uses the

old Sony 10 MP CCD sensor (ICX 493 or similar). If such a capability were available on this chip we surely would

have heard of it before.

 

Even if it were possible how would the appropriate sensitivity for each photosite be determined prior to the

exposure? It could not be done in the matrix meter since that only has 400-100 elements or so.

I fully agree with Richard. I too would like to hear a) an explaination of how sensitivity of a single photo site on a sensor can be modified and b) how that can be done BEFORE exposure. And for me it can be very technical; I've had a 40 year career in the electronics industry, including R&D and circuit analysis.

Since ADL can be turned on or off in NX, it doesn't make sense that it would work on the sensor at time of exposure.

"Do you have the reference for this claim? "

 

beyond my own notes and the notes of the few dozen journalists and people like Bjorn Rorslett in the room during the panel interview? N odoubt Niko nhas a technical paper on it and I'll see if I can find on. It is definitely a =signal processign 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. Sorry if you find it hard to believe but it is well within the realm of possibility .

 

"I've read the description of the Sony IMX 021 that's used in the D300 and there doesn't seem to be any reference to this capability"

 

Sony may be fabricating the chip in the D300 but it is to a proprietary Nikon design.

 

"It could not be done in the matrix meter since that only has 400-100 elements or so."

 

Beyond the reality that there are 1,005 RGB pixels in the Matrix meter in the D300 meter and not "100-400" and that the meter has a lot of programming based on many years of laboratory and real world testing experience behind it to interpret the range of light according to thousands (so I'm told) pre-grogrammed scenes, there is also the reality that that information can be used to very near instantaneously adjust the gain on the signal from the individual photo site as the signal goes from the photosite into the EXPEED process.

 

"how that can be done BEFORE exposure." As i said in the paragraph above te metering information is used to start a preliminary range of adjustment based on what the metering thinks is going on in the scene the processing is done during and not before the exposure process.

Sorry Ellis, but this doesn't make any technical sense. Without exposing each and every photo site on the sensor first, there is no way of knowing what light level each and every photo site receives. This means that any adjustments can only be made AFTER the exposure is finished.

 

IF the 1,005 pixel matrix metering chip were used to do any pre-exposure adjustments, then those adjustments could only be done on a maximum of 1,005 "zones" on the sensor and all photo sites within one "zone" would have the same adjustments. Apart from having a hard time to believe this, this would mean that adjustments are NOT done on a photo site by photo site basis.

 

Without knowing for sure, I bet that the "Active" in ADL refers to the exposure compensation reported by many and aimed at preventing blowing out the highlights. This exposure compensation can only be as good as the information from the matrix metering chip with the 1,005 "zones". Any variation in light levels for the various photo sites within one "zone" are invisible to the matrix metering chip and thus the exposure compensation can at best be an educated guess, even if an extensive database of possible scenes is used.

 

I'm sorry to say this about a company I admire greatly, but it looks like sofar Nikon has given everybody a pretty vague and misleading story about ADL.

 

Besides from the confusion about how ADL really works it still looks to me like a gimmick. It would make a lot more sense if Nikon gave us R, B and G channel histograms from unprocessed RAW data so we would know for sure how much room there is between maximum channel levels and the blow out values.

I have no information about what Active D-Lighting really does, but one possibility would seem to be that it

changes the amplification of the signal into the A-to-D converter on a pixel-by-pixel basis. As I understand it,

this is what happens when you change the ISO setting as well. If this is what it does, the image is captured as

electrons in the photosites the same way for all pixels, but the A-to-D conversion is different for each pixel,

based perhaps on some rough measurement of the signal stored for that photosite. All this (as well as adjustable

ISO settings, for that matter) is useful only if the A-to-D converter works with fewer bits than are needed to

exactly represent the number of electrons that might be stored at a photosite.

Changing the amplification of the A-to-D converter, if that's what is done, is after the exposure has been completed and would yield the same results as a curve adjustment. And I can't honestly see how the amplification of the A-to-D converters could be changed for every individual photo site for the many millions of them in a reasonable time. It's way easier to recalculate photo site data AFTER the A-to-D process has been completed as I suspect is the case.

Thom Hogan has a pretty good explanation of what happens, starting on page 312 of his D300 Guide and I am sure the other new guides. He does not disagree with the statements above, just explains it differently. Basically, he claims active D-Lighting in the camera will shift exposure down by 1/3 or 2/3 stop (depending on whether it is set to normal or high) to provide more "headroom" and then compensate with tone curves. The problem is this affects the RAW files as well as JPEGS including the histogram; his advice is to leave it off if the user is taking RAW files through NX since that program does it better and won't result in increased shadow noise. Just watch out for highlight saturation.

It increases my dynamic range.

Curt,

 

So does Thom Hogan agrees with Frans and myself in that ADL reduces the exposure and then modifies the tone

curve? Is he saying that the problem is that the RAW file is modified in the sense that exposure is reduced and

noise is therefore increased?

 

That will be true but to avoid blown highlights it must surely be necessary to reduce exposure sometimes? This is

exactly what I do myself on my little D40, lifting the shadows using a non-linear curve.

Richard,

Yes. The problem according to Thom is that with Active D-Lighting the exposure will always be reduced whether the image needs the highlight protection or not. Therefore he writes that you are better off monitoring the highlights yourself, reducing the exposure only when appropriate to avoid blown highlights. Now, I don't want to put words in his mouth, but this doesn't seem much different than metering for the highlights like we always did for slide film. The difference now is that we can adjust the shadows in post-processing.

Thanks for that Curt - it's now all quite clear to me.

 

I wonder if the new "auto" setting for ADL on the D90 means that the exposure will only be reduced if the matrix meter judges the contrast to be excessive?

Well, that's a good point and I can only speculate. It would seem that a matrix meter could still miss local highlights - but at least it should be somewhat better.

"Besides from the confusion about how ADL really works it still looks to me like a gimmick.

 

Since you clearly haven't used it how could you possibly know? if it is a "gimmick" it is one that works.

According to some people it should be impossible for bumblebees to fly.

 

 

 

'It would make a lot more sense if Nikon gave us R, B and G channel histograms from unprocessed RAW data so we

would know for sure how much room there is between maximum channel levels and the blow out values.""

 

So if you ever get your hands on a D3, D700 or D300, shoot NEFs and keep Active D-lighting turned off. In Adobe

Lightroom or Camera raw or your raw processor of choice make sure you have it set for a linear tone curve. That

is as close as you are going to get as all DSLR manufacturers: Canon, Hasselblad, Nikon, Pentax, Phase One,

Sinar, etc. all to one degree or another pre-cook the raw data that comes off the sensor.

I've used it and I prefer it on to being off. I haven't seen any damage to my images, as many of you have intimated here. I like the extra shadow detail and better highlight control. I took nearly 2,000 photos on a recent trip to London. I do not want to have to post-process that many photos in NX, which is a pretty horrible application speed wise. ADL is a good thing and I will continue to leave it set to "Normal" on my D300.

Apparently a moderator deleted my answer to Ellis Vener. So, let me try one more time.

 

ELlis wrote: "Since you clearly haven't used it how could you possibly know?"

 

You're right Ellis, I haven't used it because my D70 doesn't have that feature. However, I have read many reviews and looked at sample images and for all I know, and many seem to agree with me, it's nothing more than a reduction in exposure before the shot is taken to reduce the risk of blowing out the highlights and tweaking of the tonality curve like Shadows from the Shadows/Hightlights tool after the shot has been taken to restore some detail in the shadows. Looks to me such adjustments and tweaks can be made as necessary by a knowledgeable operator with more control over the results. It also looks like Nikon has been very vague and ambiguous about this function. I'd appreciate an update if and when you are able to shine more light on this issue.

Hello, I have make a small test on D300 and compare RAW output from ADL-High and ADL-Off with exactly the same aperture and shutter settings. RAW was processed in UFRaw, which does not use any special info from D300 NEF. Resulting histograms and images were found the same. Camera exposure meter show 2/3 stop below without ADL and exact for ADL with that settings.

So for me: in ADL mode camera somehow bias to underexpose, it does not make any special in-camera treatment of sensors readout, which is pity.

Conclusion: ADL in D300 will be usefull for incamera-JPEGs and, may be, for CaptureNX processing (it's tooo slow for me).

Better to set it OFF for normal use.

HI FOLKS,

 

Nikon D-Lighting is based on a patented method for dynamic range compression. Nikon licensed the patent from V.

Chesnokov (WO 02/089060, http://www.wikipatents.com/gb/2417381.html). The problem of dynamic range compression

is to map an output of say, 14 bits (input range), to a much smaller range of say, 8 bits (output range), thereby

doing better than simply clipping values which exceed 8 bits. So, if you do it by manipulating curves, you apply

the same operation to each pixel. But you can do better by choosing a different curve for each region. But

then, you have the problem to identify those regions (anisotropic diffusion may be used for that).

 

Nikon D-Lighting does an area-based dynamic range compression (citing from the patent):

 

A method of image processing comprising altering an input image using a non-linear image transform to generate an

output image, the process comprising correcting an image on an area-by-area basis to generate an output image

intensity (1',) of an area which is different to an input image intensity (1,) of the area, the output image

intensity (1',) of an area being related to the input image intensity (Ill) of the area by the ratio:

amplification coefficient= I'd/ 1, wherein the image processing method produces an output image in which the

amplification coefficient of a given area is varied in dependence upon the amplification coefficient of at least

one neighbouring area, in order that that, in at least part of the image, the local contrast of the input image

is at least partially preserved in the output image.

 

This is not simply a histogram modification, as this is a global operation. D-Lighting is more local, that is

region-based. However, there are now existing algorithms which give more pleasing results than D-Lighting.

 

AHHH - AAAnother thing: Two versions of D-Lighting exist. One that acts on the sensor's dynamic range (14 bit

for D300,D700,D3), this should be active D-lighting. Then, a post-sensor version, which acts on a smaller

dynamic range.

Active D-Lighting in D90 works just fine. If you have contrast scene, ADL help you with shadows and highlights. It is also helpfull at backlit scenes. But if you take a picture of an evenly lighted scene ( fog, diffused daylight,...), than you should turn it off. It is not wise to reduce contrast in low contrast scenes. All that is aplly, if you shoot JPEG. If you shoot RAW, than forget about ADL. In that case, you should forget all the fiddling with in-camera settings, since they have no effect on RAW anyway.

Sorry, to answer the original question: Yes, it does exactly that: raising luminance of darker regions,and reducing luminance in highlights. The art of the D-Lighting algorithm consists of (i) identify regions which are too dark or too bright, (ii) reduce the overal image contrast not too much, and (iii) avoid contrast-reversal artifacts. It is nothing but applying a tone-mapping method to a mid-dynamic range image (14 bits are not yet high dynamic range).