George Sinesios' Treatise on Olympus cameras and proper exposure

from the Olympus forum Dec 2004.

home page

Olympus C-8080WZ digital camera technical stuff

• The histogram function on a digital camera cannot help you determine proper exposure by itself. But what does the histogram graph tell us?
• First, there are no right or wrong histograms. Any variation can possibly be acceptable for a specific image and/or use depending on the 'look' desired by the photographer.
• Second, a histogram cannot indicate correct exposure, nor does it dictate if a certain image is properly exposed or not; except to possibly indicate if the dynamic range of the framed scene was captured fully with the selected
  settings (when there are no spikes at either end of the graph, nor are there any spikes (to the top edge) anywhere in the middle of the graph.
• What the histogram does show is 256 levels of Brightness starting at Level 0 (darkest on the left edge), to Level 255 (brightest on the right edge), with 256 levels for Saturation (or Amount) for each level of brightness, (starting at 0 at the bottom edge to 255 at the top edge) of the histogram's graph.
• Data shown on the extreme L/R edges of the histogram (specifically Levels 0 and Level 255) would indicate that either there is shadow information (left edge of the graph) that is not being captured with any detail, or, there is highlight information (right edge of the graph) that is not being captured with any detail.
• If the histogram is hitting the top edge at any of the brightness levels in-between, then that would indicate at least one fully saturated channel (R,G, or B) or more, and usually indicates White Balance problems or a specific colour the image sensor is overly-sensitive to with the selected exposure and WB settings.
• If the histogram is showing data (spikes) on the left-hand edge, then the ways to correct for that would be to open the Aperture (a smaller f/stop #, if we can select one), select a slower Shutter speed, increase the ISO speed and/or add more light / use Flash for the dark/shadow areas of the scene.
• If the histogram is showing data (spikes) on the right-hand edge, then the ways to correct for that would be to close the Aperture (a higher f/stop #, if we can select it), select a faster Shutter speed and/or decrease the ISO speed.
• If the histogram was showing us data (spikes) on both extreme ends of the graph, then what this indicates is that the scene has a dynamic range beyond what our image sensor is capable of capturing in a single shot; in this situation, we have four additional options:
  1. Decide to capture the detail at the shadow end of the scene and let the highlights clip (not usually recommended because digital capture usually clips the highlights in a very ungainly (ugly) fashion, use only if your subject Is the shadows and/or you are going to combine images like in #4, below).
  2. Decide to capture the detail at the highlight end of the scene and let the shadows clip (some shadow detail can usually still be brought back, but at the expense of greatly increased noise; most images (especially printed)
     will look more 'right' though with this option, than with option #1 above).
  3. Try to decrease the dynamic range of the scene by Adding light (Flash) to the scene; this will open up the shadows and compress the scene's dynamic range to better fit into the image sensor's dynamic range. The downside to this of course is that you change the scene by changing the light! There is no free lunch - use (Flash) at your own risk (to your images).
  4. Of course, there are some subjects/scenes where even #3) is impossible to even attempt, here we can try to bracket our shots (shoot #1 and #2) and combine them later in PS or almost any other image editing program. A
     tripod is essential for this technique and so is working quickly (clouds/moon/etc. can move very fast in a scene, in just a few seconds). The downside to #4, is that for almost anything other than a basically static scene, this technique is far too slow to be used in all/most normal situations (although it does give the best image quality when done properly - using a tripod and shot in RAW, and, I would recommend at least 3 shots to be combined; one for the shadows, one for the midtones and one for the highlights).
• If the histogram is showing spikes right up to the top edge of the graph, anywhere between the shadow and highlight extremes, this is indicating that one or more channels is fully saturated (Level 255) and therefore is not capturing some of the data that is in the scene. This is an indication that the WB may not be set properly for the scene, or, the sensor's response to the particular dominant colour in the scene is overly-sensitive.
• Try to correct for this situation by setting a custom WB (use the One Touch Custom WB function of the camera while pointing to a 'true' neutral white or gray area that is in the same light as your scene/subject is in), or you may
  simply have to decrease the exposure (use negative exposure compensation -EV) so that the sensor doesn't get overloaded at any brightness level.
• For WB solutions (yes, even in RAW it ensures the RGB gains are set properly and it can help you set proper exposure too): See http://expodisc.com/ 
• Shooting RAW will help in this situation by processing in PS's Adobe Camera RAW (ACR) which allows you to 'recover' blown out channels by using negative Exposure in software. This 'reserve' in the highlights that shooting RAW
  allows you is the biggest single reason to shoot RAW; you can use all 10/12 bits of the camera's image sensor and you still have the brightest, most detailed highlights possible - and at the same time, with a minimum amount of noise in the shadows too.

• The Digital ESP Multi-Pattern metering that our 'C' series cameras use compares the framed scene in our lens to a database of stored scenes to determine this 'proper' exposure. If the Digital ESP metering in the Olympus cameras is close to the Nikon Matrix Metering system (and to me, it seems like it's a less sophisticated version of it), then it has up to 30,000 'scenes' in it's database to compare/choose from.
• If the framed scene in our camera's lens corresponds to one of the stored 'scenes' in the camera's internal database (and the metering is set to ESP), then the camera will properly expose for the subject with no user intervention required.
• Chances are, the exposure the camera will suggest will be within 1/3 to 1/2 stops of 'optimal' for the specific subject/scene in P, A, or S priority modes. As long as the scene/subject is within normal reflectance values,
  then the ESP metering mode will give us at least an acceptable image file - exposure-wise - if not, dead-on exposure.
• For the 1/3 to 1/2 stop variance that is possible with ESP metering, RAW is the best way to get the exposure back to 'exact' spec's, although jpgs/tiffs can be easily corrected for slight under-exposures just as easy (or easier, some people might argue), when you need to correct for over-exposure, jpgs are almost a lost cause (compared to RAW, in my opinion).
• When you consider that the 'C' series cameras have a tendency to produce a slightly (1/3 to 1/2 stop) over-exposed image as default, then you may understand why I prefer the leeway of shooting RAW.

• But, I still haven't explained how to properly meter a subject and set the exposure.
• To properly (spot) meter a subject means to select the correct combination of settings for ISO, Aperture and Shutter speed so that we capture the tones present in a framed scene to be accurately rendered in the final image file.
• That means We must determine what is black in the scene, what is gray and what is white and make appropriate choices to ensure that black stays black, gray stays gray and white stays white.
• Why must we make these choices when the camera is in any mode other than ESP?
• In Weighted Average or Spot metering modes all camera meters assume that the scene/subject we are pointing to is 18% gray and offer/suggest exposure settings accordingly.
• If what we are pointing at with the camera is really a middle gray (18%) subject, then we can just trip the shutter and be sure that the subject will be properly exposed.
• If the object was white, the meter will still suggest to us exposure settings that will Make that subject middle gray if we simply tripped the shutter with it's recommendations - unless we use positive exposure compensation (+EV) to make the white object render white in our image file.
• We would do the opposite when the spot metered subject framed by our lens is black; the camera will suggest to Make this a middle gray object too (as it always will) with the exposure settings it suggests and unless we use negative exposure compensation (-EV), we will not get a black subject in our final image.
• Please see the following link for determining proper exposure when you are using the camera's other Metering modes - (other than ESP Metering Mode): See: http://www.luminous-landscape.com/columns/determining-exposure.shtml

miscellaneous info:

• Combine this intelligent ESP Metering with the histogram optimizing tips above and you will have the ultimate quality that the camera is capable of producing. However, all is not totally rosy with the histogram function.
• What the histogram should ideally be showing us is the Limits of the image sensor in regards to it's light gathering abilities.
• Unfortunately, of all the camera's I'm aware of, none show this most useful information - instead, they show the histogram of what an in-camera processed jpg would look like. This can be worlds apart from the actual recorded RAW data the sensor recorded. Why is this difference critical?
• First, the histogram is showing us a 'jpg equivalent histogram' (even if shooting raw ORF's). This means an 8 bit, 2.2 gamma (vs. 10/12 bit linear 1.0 gamma) White Balanced, variable Auto Leveled, variable Auto Contrast, variable Saturation, variable Sharpened, ... 'finished', image.
• Compare this 'jpg equivalent histogram' to the RAW output of the image sensor's 'true' data, and we can see 1/3, 1 or even up to 2 stops of difference (or more, in certain images) in the histograms (raw ORF's vs.jpgs)
• What this should mean to you is that the camera has a lot of 'reserve' for preserving highlight details, but also, by relying on the histogram verbatim, we are losing up to 1/2 the image sensors quality - by only using 2048 (or less) of the 4096 discreet values available in RAW for each channel.
• The best way to use the histogram is only as a (rough) guide to the image sensor's Limits, not to set proper exposure.
• If you're shooting raw, then you can/should test to see that if, in the specific shooting situation you're in, you're able to coax more out of the camera by Increasing the exposure beyond what the histogram suggests. More than likely, you will be able to increase exposure by at least 1/3 of a stop and maybe up to 2 stops or even more - depending on the scene/subject/lighting setup specifics for your shoot. What you will get in return is an image file that will be robust enough to edit it for almost any purpose, including enlarging it to wall size proportions at a quality that has to be seen to be believed.
• It may seem like the above tips only/mostly apply to RAW file processing - but that is not completely true. Although you have some leeway when shooting RAW to correct some mistakes, if you have the time to set up the camera optimally for the scene/subject at hand, then shooting SHQ/HQ jpgs will give you an equivalent quality file - exposure-wise (except, of course, for the jpg artifacts, and/or possible mis-matched sharpening and/or WB settings applied to the jpg files). If you are not going to print beyond about 10"x14" (no cropping), then these 'optimally' set up and shot SHQ/HQ files will be virtually indistinguishable from their raw counterparts.

summary:

• The histogram indicates the sensors limits. Combined with the power of the ESP metering mode and the +-EV compensation settings, it gives us a formidable basis to extract the most quality from our cameras. Only RAW will give us the Best image quality, but with a little bit of experimentation on our part (and a scene/subject that is static/patient enough), we will quickly learn the true limits of the image sensor in our usual shooting/lighting situations - and whether we choose to shoot in RAW or jpg modes we will be ensured to obtain the highest image file quality possible.