On evaluative metering, and why mostly it isn't

[Inkanyezi]

For a number of years we have had the feature of ”evaluative metering” in our cameras, not only digital, but the concept was conceived also before cameras became digital. In essence, it is supposed to mean a method of measuring light received from several different parts of a view and calculating an integrated value based upon their respective weight (importance) for the image. The goal is to achieve a correct exposure that retains highlight as well as shadow detail.

My main objection to the idea is that one can hardly expect a lump of dead metal to understand exactly what part of the intended image, as conceived by the artist, that is most important, and whether the artist would accept highlights to be sacrificed for better rendition of shadows or the other way around. Of course we have compensation settings for that task, but that overthrows the concept, as compensation is exactly what evaluation is about; the camera does not evaluate, but it integrates. The photographer is the one that can evaluate, but in order to do so, the photographer must understand what amount and direction of compensation that would be required to compensate the measurement done by the instrument.

However, considering the concept of ”exposing to the right”, implemented in the Bunsen grease spot photometer made by SEI, which still after almost a century is sung as a mantra, ETTR, such evaluation surely is possible. If we by ”correct exposure” mean an exposure that without burning out highlights contain the largest possible tonal variety of the entire scene, evaluation simply is finding out what part of the image that should be considered a highlight that reaches, but not surpasses, the highlight limit. This is the simplest implementation of the idea behind the Zone System, applied for positive processes, as conceived and described by Ansel Adams. A somewhat more elaborate version could also include identifying the deepest shadow and if possible adjust the dynamic range of the recording medium to the dynamic range of the scene. For implementing such features, it is essential that the instrument has the ability to discern the extremes within the image area. It must be able to positively distinguish the brightest highlights and the darkest shadows without blending them with different nearby tones.

When the camera measures light with the image sensor, which has several millions of light wells, each one receiving a very narrow angle of the incoming light, it is definitely possible to do such an evaluation of the incoming light. The image sensor can discern exactly what spot has the darkest light value as well as the brightest. However, if the light is measured by other sensors, the abilities of the measuring system is hampered; more so the fewer and wider apart the measuring elements are distributed. Hence a camera that uses a number of widely spaced sensors, as the autofocus sensors of a DSLR camera, cannot positively discern the brightest highlights in a scene, thus it cannot by itself perform an evaluative measurement. The worst performers in this respect would be measuring cells placed in the viewfinder system pointed toward the viewfinder screen. There is one camera manufacturer that has included a small camera in the viewfinder to cope with this problem, but in most other DSLR cameras, there is no provision in the light measurement system for identifying highlights.

And there I reached the crucial point of why evaluative measurement isn't. The main issue is that it cannot be evaluated if it is not sorted out and when the data is not registered. Compact cameras and mirror-free system cameras have the ability to find the highlights, and there are provisions for evaluation in several of them. Many of those cameras can present blinkies for highlights as well as shadows, so the photographer may easily evaluate the scene, and the process could be automated. However to be creatively useful, the photographer needs a basic understanding of what the system does. If the system is badly conceived, as well as poorly explained, it is very difficult to find out what it really does. From an operational point of view, it is essential that the photographer has a basic understanding of the system and how it will impact the images. One could argue that a simple presentation system as the Zebra mode of CHDK or the under/over blinkies of Olympus, might serve a creative photographer better than a fully automatic system for evaluative metering. The latter does exist in Nikon DSLR cameras, Active D-Lighting, and photographers differ about its usefulness and practical value. Many regard it as utterly useless, but some would not want to be without it.

Here is where I come to an issue that was differed about in a recent thread, 10 kinds of photographer., the ability of Canon DSLR cameras to evaluate the viewed image, to effectively perform evaluative measurement. There are discussions about it in Canon forums, and there are opinions about how well it functions. Canon even has it on their own learning environment on the Internet, and they explain rather clearly, but also very fuzzily how it is supposed to work. However the explanation they give has several weak points, and the pattern outlined in their sketches does not represent DSLR, but the screen of compact cameras. They do not present any evidence or algorithms of what is engineered into the system or what it achieves. The whole explanation is a bit like a marshmallow. No hard data, nothing to hold on to. However, when I scrutinise it, it seems as they measure light for evaluation with the AF sensors, which are few and widely spaced. There is no way the AF sensors could discern highlights, and the examples that are given clearly show that it is the chosen AF point that carries the largest weight, and that it is calibrated to middle grey. Hence if you focus on something bright, the camera will under-expose the image, and if you focus on something dark, your image will come out over-exposed. Thus the behaviour of the system is inconsistent and far from intuitive. It forces the photographer to basically understand the Zone System and apply it to the chosen focal point in order to get a correctly evaluated exposure. Hence the metering is not evaluative in the sense that the camera evaluates, but it demands of the photographer to evaluate the chosen focus point in order to expose correctly for it. For a dark focus point, negative compensation is needed, and for a bright focus point positive, just as with spot metering.

I took the time to pick three references, two from Canon:
http://learn.usa.canon.com/resources...ng-Zones.shtml
http://www.learn.usa.canon.com/resou..._article.shtml
The third is from a reviewer:
http://www.completedigitalphotography.com/?p=1810

[Mike Buckley]

it is essential that the photographer has a basic understanding of the system and how it will impact the images...the [Canon] examples that are given clearly show that it is the chosen AF point that carries the largest weight, and that it is calibrated to middle grey.

There have been at least a couple Nikon DSLRs, one bieng the D80, that have been demonstrated to work the same way, at least to the point that the luminosity of the scene where the AF point is being applied can dramatically affect the resulting exposure. I don't remember if that's because the point is calibrated to neutral grey.

The issues being discussed about evaluative metering have never bothered me. One reason is that I presume that the term, "evaluative metering," is conceived or at least promoted by the marketing folks of the large corporations. That in itself causes me to be skeptical at worst and cautiously optimistic at best about the accuracy of the term. The second reason is that I always check the histogram after releasing the shutter. Doing that allows me to verify whether the camera exposed the scene in a way that meets my goals.

[Inkanyezi]

What first set me on the track is that most cameras in fact lack the actual hardware to perform the function of evaluating. Even though it is present in Nikon cameras, it should be understood, that those fields that are measured are rather large, so that each one of them may contain not only a single highlight or shadow part of the scene, but a mix of them. Neither of my cameras can be trusted to actually evaluate lighting (Canon G7, Lumix G1, OMD, EOS 10D), but the G7 and the Olympus have both histogram and highlights/shadows available in real time, the G7 with CHDK. The Lumix has a kind of real time histogram, but I don't know what it's good for, and the EOS will show it only after the picture is taken. There is a function in the Lumix that purportedly stretches dynamic range, but I haven't figured out how it works or is supposed to work.

The Wikipedia article on measuring methods is rather straightforward, but not displaying scepticism against marketing blurb: "Many manufacturers are less than open about the exact calculations used to determine the exposure."

I have figured out how the Nikon system works in those that have Active D-Lighting, as well as those with only D-Lighting, and the "Active" really tries to adhere to the idea of exposing for highlights while D-Lighting developes for the shadows. Canon marketed something that was said to be similar in the EOS 40D, although it was not implemented, it only changed the appearance of the zeroes following 2, changing 200 to 2oo. Later models seem to reduce exposure however and change the curve in a similar fashion to D-Lighting. It is actually fairly simple to do the latter operation in PP with the Curves tool.

No manufacturer afaik has released any white paper on what those elusive functions actually do. Nikon seems a bit more open than other manufacturers, but I don't trust that talk about thousands of example images stored in memory to compare, and I doubt that it would be possible to make a suitable algorithm for the chore. My guess is that it tries to set exposure for the brightest part it can find with the rather advanced measuring system, and then it develops with a curve that is lifted in the shadows.

[Manfred M]

Urban – A lot of the points you have made are the same things I have been wondering about. Some functions in my DSLRs have resulted in tangible benefits in my photography; which I define as an increase the percentage of good, usable images from a technical rather than aesthetic (compositional) viewpoint. I’m afraid that I really have not seen any tangible improvements in my exposures, versus the technology in my 30+ year old film cameras.

We have not seen any white papers from the manufacturers explaining how this technology works, and anything we do see comes from marketing literature. I do believe that the camera manufacturers can back up their claims; otherwise there would have been legal challenges. That being said, just because the statements are correct, that does not mean that there are any tangible improvements to the end user.

I still have to use exposure compensation to get good night shots, backlit scenes or snowscapes, so these rather obvious situations seem to be beyond what the technology can handle. I still get blown highlights or loss of shadow detail when relying on the camera’s light meter, but seem to be able to fix them by reviewing my exposure using the camera’s histograms.

The part of the “evaluation” that appears to be missing is that the camera still assumes an “average” scene with 12% or 18% gray scale, depending on who you talk to. Sophisticated technology that relies on a fairly crude baseline seems to be the underlying reason for the underwhelming results….

[dje]

One of the things I like about photography is that there is always something new to learn. My lesson for today is that in what Canon call Evaluative Metering Mode (hows that for diplomacy), there is a link between the metering sensor and the AF sensor point(s) selected. This is described in the second Canon link referred to above. This also explains the statement in my 600D handbook which says that with evaluative metering, when the shutter is half depressed, the exposure as well as the focus is locked in.

My take on this is that for evaluative metering, the AF sensors selected are not used for actual metering but are used to tell the metering sensor and software which parts of the image to give prominence to ? See this link

Anyway, thanks for helping in the progression of my education !

Dave

[Colin Southern]

Sorry, but I think this is being waaaaaaay over-thought.

In short, ANY kind of metering - in essence - simply makes a decision based on programmed parameters. Photographers - knowing the principles (but not the details, because that's proprietary information) can then modify that decision (if they wish) to achieve the result they want. That's it. Nothing more, nothing less.

I once suggested to Canon that they include an ETTR mode, but since then I've come to realise that it just wasn't a real-world idea (a) things like specular reflections would totally stuff things up (b) it drives the sensor (or more likely one channel) into a non-linear region where you you end up with weird colour shifts in the highlights that are nigh on impossible to correct (c) it makes the midtones harder to get exactly right (d) it's just not needed in the vast majority of every day scenes (considering the camera already captures around 64 times more luminance range than most monitors can display) and (e) it's nothing that can't be achieved by adding approx a couple of stops of +EC (assuming base ISO). So - please forgive me Canon for making such a silly suggestion, for I didn't understand at the time -- but now I do.

I'm using evaluative metering 99% of the time - and 99% of the time it works just as I would expect it to. If it ain't broke, don't fix it. And it ain't broke.

/out.

[Inkanyezi]

I can definitely see very tangible results from the new technique that is used versus the old. It mostly does not change much substantially, but it achieves the goal in a more consistent way, and the difference is most noticeable when conditions are tricky, particularly high contrast scenes. But even though I have access to a few reflex cameras (D7000, D800, D90, EOS 10D, 40D, 7D), the only system camera I actually use presently is the OMD E-5. Its highlight and shadow warnings, as well as live histogram are a great help when determining exposure, and those conditions that earlier could include guesswork are now resolved easily by seeing in the viewfinder or on the screen exactly how tweaks work out. The width and position of the histogram indicates whether there are unnecessarily large margins either side, often enabling a faster shutter time with higher ISO setting when I won't need very high dynamic range. Blinkies tell me when I would get blown highlight or sooty blacks. And I have exactly the same information when using my PowerShot G7 with CHDK and Zebra mode.

None of the reflexes I have tried provides the same ease of evaluating exposure, and only the Nikons are fairly reliable in high contrast situations. With the Canons, the guesswork is just the same as it was with my old cameras using film, Contax RTS, Olympus OM1, Topcon Uni and Unirex (the latter two actually used professionally for using flash as they have leaf shutters). But of course with digital, the image can afterwards be analysed on the display. In my view, it is a highly valuable feature to be able to do this scrutiny also before pressing the shutter. It is particularly valuable in order to choose optimal setting of ISO for getting fast shutter times that are desired, still retaining reasonable dynamic range. The live view aids for exposure clearly show the match between the sensor's dynamic range and that of the scene.

[xpatUSA]

. . . .The Lumix [G1] has a kind of real time histogram, but I don't know what it's good for . . .

The Panasonic Lumix DMC-GH1 also has the real-time histogram that displays at top left of the LCD - and in the excellent EVF, if you use it. The GH1, as we know, is a G1 with video added.

Perhaps it's good for evaluating the exposure right up to the moment of pushing the button? Or using while filming a video?

[DanK]

I had pretty much the same reaction as Colin. Most of the details of how the algorithm work won't affect how I meter. If a scene is clearly hard to meter, I use a spot meter to evaluate important regions, e.g., the brightest that I don't want to blow out, and use an informal zone system, or I use the spot meter off my palm and open up a stop. I spent years with a Canon FTb, so using a spot meter is second nature to me. When the lighting is not tricky, I often use evaluative metering, and like Colin, I usually find that it is pretty close. When it isn't--well, that's one nice thing about digital. in the film days, I worried more. Now, I check the histogram, and if the initial metering was wrong, I just change it. The one part of the discussion above that will affect what I do is the fact that with Canon systems, evaluative metering is affected by the choice of AF point. I so often use center-point-only that I had not realized this.

[pnodrog]

To a certain extent this discussion is becoming irrelevant as more cameras incorporate scene recognition software to assist in optimizing both metering and focusing required. As sensors improve in dynamic range I think in the near future exposure evaluation will become of far less concern than accurate focusing. We are not far off the point where the exposure as determined by the camera is sufficiently accurate that it is within the range of PP to achieve a satisfactory result for all but the most extreme exposure conditions.

Currently and in the immediate future I see no sign of being able to tweak focus to any practical extent in PP. I suppose sometime in the future we will be able to take shots in stereo with tremendous depth of field and latter adjust the apparent focus completely in PP just using our smartphone. Alternatively auto bracketing of stacked focusing.

[dje]

The one part of the discussion above that will affect what I do is the fact that with Canon systems, evaluative metering is affected by the choice of AF point. I so often use center-point-only that I had not realized this.

Same goes for me. It doesn't seem to have caused any real problems though. One thing that I will need to be mindful of though is when I direct the camera at a point I want to focus on, half press the shutter and then re-compose for the actual image I want. With evaluative metering (which I normally use), this locks in focus and exposure prior to the final composition so under certain circumstances the exposure may be out if it is based on a different composition. Always check the histogram afterwards !

[Mike Buckley]

I had been using center-weighted metering on my film camera so I used it on my first digital camera. A friend told me to try evaluative metering and I couldn't believe the improvement. It had such a lasting impression on me that I remember the exact scene I was photographing at the time. Like Colin, I use evaluative metering 99% of the time.

[Glenn NK]

Same goes for me. It doesn't seem to have caused any real problems though. One thing that I will need to be mindful of though is when I direct the camera at a point I want to focus on, half press the shutter and then re-compose for the actual image I want. With evaluative metering (which I normally use), this locks in focus and exposure prior to the final composition so under certain circumstances the exposure may be out if it is based on a different composition. Always check the histogram afterwards !

I have both my Canon bodies set up to use the * button for focus (the one under your left thumb) - the shutter button only releases the shutter.

http://photography-on-the.net/forum/...k+button+focus

Glenn

[dje]

I have both my Canon bodies set up to use the * button for focus (the one under your left thumb) - the shutter button only releases the shutter.

http://photography-on-the.net/forum/...k+button+focus

Glenn

Thanks Glenn. Yes the use of the * button for focus gets around the problem. I have considered using this before but was concerned that on occasions, particularly with hurried shots, I might forget to focus altogether. On the other hand, if a discipline is established, then it has the advantage that it makes you think about what you want to focus on every time you take a shot. I'll have to think about using this again.

Dave

[jcuknz]

The Panasonic Lumix DMC-GH1 also has the real-time histogram that displays at top left of the LCD - and in the excellent EVF, if you use it. The GH1, as we know, is a G1 with video added.

Perhaps it's good for evaluating the exposure right up to the moment of pushing the button? Or using while filming a video?

The FZ50 has/had it before[?] the GH1 I just switched it off from cluttering up the viewfinder last night Sorry to intrude on a serious and ernest discussion.

[Inkanyezi]

The FZ50 has/had it before[?] the GH1 I just switched it off from cluttering up the viewfinder last night Sorry to intrude on a serious and ernest discussion.

Shouldn't be sorry, your comment is as valid as any other. I also uncluttered the screen or viewfinder by shutting it off on my G1, as it was very large, not very helpful, and made framing much more difficult as it covered a large part of the screen. On the G1 the real life histogram is very different from other histograms we are used to; it's just weird. I don't know what it is like on other models.

[xpatUSA]

The FZ50 has/had it before[?] the GH1 I just switched it off from cluttering up the viewfinder last night Sorry to intrude on a serious and ernest discussion.

Yes, my post was a little tongue-in-cheek, too

I bought the GH1 mainly for Wifey to shoot her eBay stuff, instructing her set it to "intelligent" and watched with cosmic amazement as the LCD filled up with little rectangles, something green appeared and the camera went "beep", as if to say "OK, dumb@ss, you can press the button now " . . She loves it! Gets good pics every time.

Meanwhile, shaking my head sadly, I head back to the SD10 house-brick with it's big chunk of real glass viewfinder, it's weighty FF macro lens, and continue taking tens of manly shots in full manual and center-spot metering just to get one good shot . . .

[Glenn NK]

Thanks Glenn. Yes the use of the * button for focus gets around the problem. I have considered using this before but was concerned that on occasions, particularly with hurried shots, I might forget to focus altogether. On the other hand, if a discipline is established, then it has the advantage that it makes you think about what you want to focus on every time you take a shot. I'll have to think about using this again.

Dave

Dave:

Oddly enough, I've never forgotten to focus - maybe shooting manual lenses for so long trained me to look carefully through the VF.

Generally once people have tried it, very few seem to go back to the shutter button to focus (which I suspect was designed for photographers just starting out - just like the P setting. ) at least that's the impression I get from reading dozens of posts on POTN.

Glenn