Showing posts with label Histogram. Show all posts
Showing posts with label Histogram. Show all posts

Sunday, 10 June 2012

Exposure, the Metering Pattern and the Histogram

What is correct exposure?

What criteria do you use to assess correct exposure? Is it reading the histogram? Is it setting the camera to P A S and flipping to M to “match needle”? Is it buying this pro looking external exposure meter, then transferring the settings to the camera on M? Is it the way the photo “comes out?

There is no correct exposure. There is contextually suitable or relevant exposure. That means if you are wanting the face just right, it is. If you want the sky, it is. If you want that bright sky and that darker ground just right, well that’s being just…. greedy.

  1. The exposure histogram describes the brightness levels of the scene in numbers.
    1. It’s not a visual assessment. I’ll say it again, its not a visual assessment.
    2. It does not show whether the exposure is correct (see point 1 – there is no correct exposure). you can have what looks like a “good” histogram, whatever that is, but the face you want is contextually too dark or too light.. ditto with the sky and the earth.
    3. It can show that you are clipping highlights (255) or the shadows are going too dark ( 0 ). that does not mean you photo is “wrong” – it just shows that technically in numeric terms, you have lost digital data – that does not mean you have lost visual aesthetic.
    4. The averaged histogram may be misleading in technical terms. because your separate r g b channel histograms are different and one of them may already be clipping whilst the averaged histogram may not indicate that.
    5. There is more than one histogram for the same scene. Yes. There are:
      1. the Liveview histogram which is quickly computed so that the camera’s Liveview does not slow down. It is the stats BEFORE the shot is captured.
      2. the exposed photo JPEG histogram – this is the stats AFTER the shot is captured.
      3. the invisible and yet-to-be-deciphered raw file histogram. The reason why it is not visible can can’t be displayed is that the raw file has not been processed against your preferred set of contrast, saturation gamma curves. And if we did apply our preferences, it would no longer be the raw histogram, it would be the processed photo histogram.
    6. You might consider the highlight / shadow blinkies as an alternative to the histogram. But, the highlight and shadow blinkies give you even less information than the histogram – it can distract you by corrupting your perception of whether your exposure is contextually correct by painting your LCD with big blobs of red.
  2. You will often encounter naturally lit scenes that exceed the dynamic range of a single shot, even if you shoot raw. You will have to turn the tables (Star Trek’s Kobayashi Maru) by any trick – come back another day / season / time, stand with the sun behind your back, carry out exposure blending from multiple exposures, big a bigger sensor, shoot film).

Generally, we have:

  • Spot Metering
  • Centre Weighted Metering
  • Evaluative / Matrix Metering

Spot Metering

It tries to make anything you point at, 12% grey in tone. Point it at a white shirt and it will try to make it grey. Point it at a black shirt and it will try to make it grey.

That’s not a general use, point and shoot metering pattern. Some people figure they can use it like this, well and good for them.

Centre Weighted Metering

Centre Weighted covers a central area – yes, around the centre AF spot. The area may be hard edged. Or not. On optical viewfinders you may not be able to estimate how big the area is.

It has sufficient area that if you point it at something, it will average that bit of face, that bit of shirt and that bit of background to 18% grey. That might be just what you want – the face might turn out alright.

It has a small enough area to exclude the bright sky or that dark table shadow from dominating your camera’s exposure proposal.

There is also on some cameras, a spot biased centre weighted choice.

Matrix / Evaluative Metering

You could get the meter to measure the whole scene and take a dumb average. If you do that, some bright sky could “pull” the meter reading quite unfavourably towards a darker exposure.

So the camera makers came up with Matrix / Evaluative. Some cameras even have computationally enhanced evaluative (with elements of Artificial Intelligence, database statistics on image scene modes, face detection etc…)

This is the metering pattern (well it is not a pattern, it is a tremendous state of the art calculation) that the camera makers put a lot of their skill and knowledge into. If there is one thing they can create to help the millions of customers who just want to aim and press the trigger, this is the epitome of technological prowess.

So far on my cameras which are not leading edge nor are they the most expensive, this form of metering still can’t make it happen for the majority of my shots. Yet. And I can’t predict in which direction and by how much it will bias the camera’s setting since it is using some dynamic logic, not a standard pattern.

The Bottom Line

I have met a whole bunch of amateur photographers of various experience levels. And encountered even more on the web and in forums. Using my preference – Centre Weighted and the subjects I shoot, let me put forward some advice:

  1. Each country (e.g. Australia, Malaysia), each season (blue sky summer, gauzy overcast winter) and each time (Golden Hour, mid day, etc..) merits different consideration and care.
  2. There is no set-and-forget. And there is. Depends on your style and what you approve of and accept. You can accept that, for example, Street Photography of strangers cannot be controlled and is an unrepeatable point in time so any shot is good, even a technically imperfect shot. Or you can put on your grumpiest attitude and criticise the hell out of every landscape / panorama shot.
  3. These are modern cameras. Centre Weighted should be predictable. I didn’t say what you want (who knows what you want?). But it should be predictable. On a “standard” scene “what is standard?”, it should either be pretty close to good, or within 0.7 dark or light. If it is too dark or light, remember how it behaves and set the bias semi-permanently. It is semi permanent because there is no standard scene.
  4. Learn to identify several types of scenes. Why? Because that is what the camera makers are also doing for point and shoot people – that mechanism is called SCENE Modes. I’m NOT saying you should use SCENE mode. I’m saying you need to identify the scene you are pointing at, test it out beforehand how your camera’s Centre Weighted Pattern predicts and develop your own exposure biases so that they become instinctive and intuition.
  5. If you are using an EVF / LCD mirrorless camera, it’s even easier. Set your Liveview to exposure simulation mode (they’re already defaulting to that) and point, assess the darkness / brightness of the scene and flick your Exposure Compensation Dial. Just like that.

Further Reading

Friday, 6 May 2011

The Four Stages of Exposure Awareness

Stage 1: The first thing that newbies learn about is that there is an Exposure Triangle. Some Peterson guy is said to have wrote about it in a book. I haven’t read it. I’ve seen his videos. Maybe his intentions are good and he knows what he’s doing. But a heap of newbies don’t “get it”

Stage 2: Eventually it dawns on people that the Exposure Triangle has a Fourth Side.

Stage 3: After rummaging around, comparing effective techniques of whether to use P A S M or figuring out which metering pattern is better – Evaluative Matrix vs Centre Weighted vs Spot vs the classical Sunny 16 rule vs Interpreting the Histogram vs ETTR and asking themselves where the hell they put the white towel / Kodak 18% Neutral Gray Card or the XRite thingamajig, someone mentions that Adams chap who wrote about the Zone System. And bang! Smack on the head. There is no Correct Exposure. There is what the camera measures as an instrument and what the artist (you) choose to convey and interpret. The two are not and do not have to be the same thing

Stage 4: So far, so good. People are shooting decent shots. But they’re not spectacular. Like those gorgeous smooth skin tones and sharp, clear irises of the girls in the portraits. And so on. So we ask, how on earth does so and so get this shot with his iPhone but we can’t and we’ve almost spent as much as a Nikon D3s? And the penny drops. We can’t. If we REACT to the scene. Often times, the pros don’t react, they’re pro-active. They light up the scene the way they want. Or gain a vantage point if they can’t control the light. And having done their utmost to light the scene well, they touch up with Photoshop. Delicately and Emphatically. Not the other way around.

Wednesday, 18 February 2009

What is Correct Exposure Part 1

his is an oft repeated question by newbies to photography. An oft repeated reply by people who have “seen the light” is that the “correct” exposure for a scene is what you want it to be.


We often point to the exposure meter in the camera (classic film SLRs) or histograms in the DSLRs. and we tend to say, you must try not to “burn” highlists (overexposure) by avoiding histograms peaks pushed flush against the right edge. And you must not push histogram peaks against the left edge otherwise you get inky dark shadows that have no detail (underexposure). Heck, even some photo competitions state one of the rules is that you must not submit a shot where the Photoshop eyedropper goes 0 or 255.

Well, that’s generally true. Except that often, you can’t avoid one or the other or both. That’s because, regardless of the medium – negative film > transparency film > 24x36mm sensor DSLR > APS-C DSLR > Four Thirds DSLR > small sensor compact, the natural outdoors or indoors harsh scene will have a Scene Brightness that exceeds the Dynamic Range handling ability of the camera. I was musing on that when I compared my photos, with technically correct histograms against someone else’s photos where there was so much “pop” in them. Having a technically correct histogram does not make the photo visually interesting.

That’s because the histogram is only a two dimensional report of scene and subject brightness (well, maybe two and a half because you can have separate R G B histograms as well). The histogram does not tell you:

  1. which part(s) of the scene is causing the spike(s)
  2. about what curves, shapes and visual geometry make the subject interesting
  3. whether the face or interesting part of the scene is “correctly” exposed – can you see the detail, texture of of the face and so on.

In a recent article, The Online Photographer describes and discusses Dynamic Range and even Local Contrast. It’s a good read.

So, what is Correct Exposure? Tags: ,
Reblog this post [with Zemanta]

Thursday, 8 January 2009

Interpreting The Histogram

Lots has been written about the Photographic Histogram. Many people still don’t understand what it is and what it does. Here are my notes:

What the histogram is

  1. The histogram is a coarsely scaled and over small 2D representation of the tones. Some cameras have separate RGB histograms.
  2. The histogram has two edges, the left and the right edge of the bounding box
  3. If it looks like a peak has a slope extending past a box edge it means that there is some technical data that has been lost / will be lost regardless of RAW.
  4. RAW has a higher latitude because the contrast curve is not "baked in" and the camera histogram may not show the RAW histogram, it may show the histogram of the embedded JPEG in the RAW. So, you infer that you have half a stop or more of latitude even though the in-camera histogram is showing that there is a truncation of the slope or tail of the peak.
  5. With cameras with LiveView, the histogram is a predictive forecasting tool - it tells you before you press the trigger, what could be captured in tones. Shooting RAW or shooting a bracket of 3 or 5 shots is *NOT* a predictive tool - you have made the shots, you can retire without seeing the shots or you may take the time looking at the shots in camera - either way, you have lost oppurtunity or time. This lost time may be ok for a landscape shot but not useful for a quick assessment of a marching parade or anything in motion.
  6. Some cameras don't have histograms
  7. Some cameras don't have a good EVF or an LCD (e.g. my Kodak P880). This causes me to underestimate the visual quality of the image.
  8. Some cameras have an over beautiful EVF or LCD (e.g. the newer Canon G models and reportedly the Panasonic LX-3, the Nikon D90 DSLRs set....). This can cause people to overestimate the visual quality of the image.
  9. Some cameras don't have auto bracketing e.g. the Nikon D60.

What a histogram IS NOT

  1. The histogram does not measure the visual quality of the shot. It is after all a crude 2D graph of tones.
  2. The histogram does not tell you which subject or part of the subject constitutes the tail or slope of the peaks that is being cut off.
  3. There is no "correct" histogram - the idea of "correct" does not apply because of points a. and b.
  4. The histogram does not tell you how to compose the shot.
  5. The histogram does not tell you what is wrong with the shot. It does not advise you that you are shooting into the sun, it does not tell you that you could enhance detail and micro contrast by changing your position and your lighting angle. It does not tell you that your metering pattern is inappropriate. It does not tell you that you are metering for a 12% grey target when you are facing white snow.
  6. The histogram does not tell you clearly how much to compensate in EV - i.e. how much EV to dial in to move the peak and thus the tail of the hills. You can do some experiments and gain some understanding by dialing in EV and shooting a test subject. And watching the histogram move horizontally. You do that as homework, not on the day.
  7. The histogram does not tell you how much tail to chop off or to force into the bounding box - that's your choice and you have to visualise using whatever method you understand (zone system, experience with the subject etc....). This is a human, visual assessment, certainly the computation and artificial intelligence is getting smarter but the histogram isn't - the histogram is a crude 2D graph.

If a person does not understand how to use the histogram, there is nothing stopping the person from doing a bracket of shots or interpreting visual quality by looking at the LCD screen or the EVF.

The histogram is an informative and useful tool above the:

  1. It's sunny so I use f/16 1/100 @ ISO 100
  2. Oh, the meter says it is EV 17 @ ISO 100
  3. Heck, I'l just dial RAW and shoot a bracket of 5 shots at .3 intervals, one of them should work.

If a person does not want to use the histogram and has some magic recipe that works for them, just do it.

Reblog this post [with Zemanta]

Sunday, 13 July 2008

Epiphanies on Exposure

The Pope is in Sydney, leading the celebration of World Youth Day. My epiphany doesn't relate to that event but what Kurt Petersen took the time to find out and tell us.

Spot-Hi meters at ~215-220 on the histogram and is equivalent to Spot +2EV. Use this on something you want to be the brightest part of your photo.
Spot meters at ~115-120 on the histogram
Spot-Sh meters at ~15-20 on the histogram and is equivalent to Spot -3EV. Use this on something you want to be the darkest part of your photo.

I knew what Spot Metering does - point the center of your viewfinder it at something and the camera reports the Exposure Value you set on your camera (or the camera sets for you) to make that subject's tone equivalent to 18% Gray (or was it 12% Gray? - doesn't matter, it's about Gray).

By using Spot Metering in a Point and Shoot snapshot way, you get awful results - you may not be pointing at the correct area to make 18% Gray and you may be more concerned about not burning out the bright clouds or rendering some darker coloured ground object without noise rather than anything related a midtone.

So, if you realise that you don't want things to be Gray is an ephiphany. You actually want the opposite - to render bright highlights and dark ground objects in a pleasant way with detail,  colour and not much image noise.

That's where the ESP or "smart" biased averaging matrix comes in with these modern cameras. They detect all the tones in the metering areas and calculate a "middle" EV based an optimum chance of getting it right. But how do they do that? They don't know whether you want perfect cloud and sky tone at the expense of darkened ground, neither do they know you want the person's face or detail in grass in preference to the sky. The camera makers solve this by doing a lot of research on actual photos in the field, generating stats on scene types, metering patterns, capability of the recording sensor, people's choices in those photos and so on. So they come up with a "smart" calculation to recommend to the camera. That's the best they can do, but that is NOT YOU - You're the end of the line, the "moment of truth" - you and the scene and the camera are where it comes down to right at that moment you press the trigger.

So ESP or "smart" biased averaging can and will be wrong at times.

This is where Spot-Hi and Spot-Sh come in for the Olympus E-510. If you set the Spot-Hi to the Automatic Exposure Lock (AEL) button action, pressing the AEL button will activate Spot-Hi instead of ESP for that shot. You point it at a bright patch that you don't want to "burn" - for me, that's the clouds in the sky. You recompose and shoot. Sometimes, this will give a result similar to using the ESP pattern. Sometimes, it will darken the ground and give you a good sky tone. If you ensure that your ISO is as low as practically possible, i.e. ISO 100, then the amount of noise you generate by lifting the dark tones in post processing will be as low as possible.

What about Spot-Hi? Well, there is only one AEL button on the E-510, so I dedicate the Fn button to MyMode-1 and in this MyMode, Fn invokes Spot-Sh. It won't work like the AEL button, but if I point the camera at a dark ground object and don't re-frame, just shoot, then I am pretty sure that this dark ground object will not be inky, unrescueable black.

Below are some shots I've done using several approaches. They're at the Blackburn Lake area and are of fairly high Dynamic Range oppurtunities.

Monday, 7 July 2008

Photographic Terminology Part 2

ETTR - Exposing To The Right - is about setting your exposure such that your histogram right toe touches but does not get chopped off by the vertical right axis of your histogram. This way, you capture the maximum amount of light that your digital sensor and in-camera electronics processing can take. This gives the highest Signal to Noise ratio for your photo.

Above is a 2 MP Nikon 775 shot. The light was "right", you can see good grass colour, blue sky and clouds. Capturing a both sky and ground tones with this camera can be quite difficult - the sky usually gets burnt to white.

Stu Maschwitz of the Prolost blog argues that there is no one blanket philosophy for exposure (dead link).  If your real scene is of such high dynamic range that it the histogram is so wide that it spans and exceeds both the left and the right axes, then, you're between a hard place and a rock. You either call it a day or come up with a James T. Kirk solution to the Kobayashi Maru.

Luminous Landscape is a well known reference website for photographic detailers. They have an article on ETTR. Emil Martinec writes a scholarly discourse on Image Noise, Dynamic Range and Bit Depth in Digital SLRs - his webpage is at the University of Chicago.

On the other hand, spotted at the DPREVIEW Sony forum ...