In the 3 previous articles we talked about exposure and how to adjust shutter speed, aperture and ISO to obtain correct and creative exposure.
In this final article on Exposure in Digital Cameras we're going to look at the affects of under and over exposed images and subsequent digital noise produced and how to use the histogram to help avoid under and over exposed areas in photography.
In digital photography when the sensor has not been exposed to a proper amount of light, by either under or over exposure, the image digitized by the Analog to Digital (AD) converter will not produce all of the data. This is due to a phenomenon called clipping. It is caused by an under or over load condition that results in the incoming data stream not being picked up by the sensor. In an under load condition, the sensor is not sensitive enough to pick up the faint light. In an over load condition we bombard the sensor with more information than it can resolve. In the case of the under loaded or under exposed image, the sensor is trying really hard to convert the analog light into digital code but all it picks up is background noise.
An analogy here would be trying to record whispers across the room on one of those old time cassette tape recorders with the built-in microphone. If the whisper is not at a level high enough for the microphone to hear then all the microphone hears is background noise and it records that. So you hear the hum of the tape recorder motor, the hiss of the cassette tape and any audible background noises in the distance like cars driving by, birds chirping, and kids playing. The camera sensor works just like the microphone in that if there is not enough image date available it will pick up background electronic noises of the camera and sensor assembly itself. If the image is properly exposed then the image data will override the background noise and even though the background noise is there you can't see it and in the case of the cassette recorder you can't hear it.
In over exposure the opposite thing happens. The sensor is bombarded with bright light and the sensor's AD converter is kept pegged at max voltage. Max voltage for the AD converter means pure white.
So as a result the under exposed parts of the image have background noise and the over exposed areas in an image are pure white.
If we try to raise the levels of the underexposed image post processing to brightening the under exposed areas all we do instead is amplify the recorded digital noise. Since there was no image detail recorded to begin with there is none in the post processed brightened image.
If we try to lower the levels of the overexposed image all we do is change the pure white to shades of grey. Since there was no image detail to begin with there is none in the post processed darkened image.
Now, knowing this information we can understand why there is more noise in the under exposed exposed areas of the image when we turn up the sensitivity of the camera's sensor (ISO). Just like if we turned up the microphone sensitivity of the cassette recorder it will only record the background noise even louder. In other words if the levels of the sound or light we are recording are lower than the background noises of the recording instruments then turning up the sensitivity is not going to help record them any better and only records the noises louder.
Now there is nothing wrong with adjusting (within reason) the sensitivity of the camera's sensor if we are recording levels that it can record, in other words a properly exposed image. It's one of the settings that allows us to have greater control over our creative exposure.
The problem usually lies when we capture an under or over exposed image. The problem lies more with low light conditions and less with bright light. The camera can usually be adjusted in such a way to block the extreme light from the sensor, either by fast shutter speed or small aperture or even a neutral density filter or two attached to the lens. However the opposite is not the case. In very dim lighting you can only increase the ISO so high while the fastest lens can't let more light in than there is available and the shutter can't be open too long before the sensor starts to pick up the background noises due to multiple hits on the same sensors. Fortunately we live with this handicap of not being able to photograph in very low lighting or we bring our own lights to the table to compensate.
So how do we know when an image we just captured falls within the camera's selective recording range? We use the histogram of course.
The histogram is a great but often misunderstood tool that most if not all digital cameras make available to us. The histogram is a rectangle graph divided into 4 or 5 vertical sections depending on the camera. The K10D and most other digital cameras have 5 vertical sections. The left side of the graph is the shadow area. The extreme left edge indicates the sensors noise threshold and any data that is butted up against the left edge is underexposed and subject to noise. The right side of the histogram graph is the highlight area. The extreme right edge indicates the sensors max voltage level or pure white. Any data that is butted up against the right edge is over exposed and recorded as pure white. In either case the extreme edges is where the image data has been cut off or clipped and no amount of post processing brightening or darkening will reveal it. All you will do is amplify the recorded noise. Any data between the left and right clipping edges for creative exposure sake we'll call properly exposed data.
So what to do if the data is underexposed or bunched up against the left edge of the histogram? Increase your exposure a stop or two until there is a gap between the data on the left of the graph and the left clipping edge. What to do if the data is over exposed or bunched up against the right side of the histogram? Reduce your exposure a stop or two until there is a gap between data on the right side of the screen and the right clipping edge.
What to do if in the rare occasion you are trying to capture an image that has data clipped on both edges of the histogram? The quickest and easiest solution would be to reduce the exposure until the right side is no longer clipped and leave the shadows where they lie. An image looks better when the highlights are not blown out even though there is no detail in the deepest shadows. And as long as you don't adjust (amplify) the image levels to try and bring out the shadow details you won't notice any noise. If for some reason you need the shadow details then of course do the complete opposite and properly expose for the shadows not to be clipped on the left.
The alternative solution, and a bit more extensive work in post processing, would be to take several exposures ranging from properly exposed shadows to properly exposed highlights. This could be as few as 2 images (3 is the usual minimum amount though) to 5 or even 7 images of varying exposures. This is called exposure bracketing and the K10D has an auto exposure bracketing feature where it will expose 3 or 5 images at exposure levels that you set. If you need more than 5 bracketed images (even under extreme exposure conditions 5 will usually suffice) then you have to manually take the images. A tripod is required when doing this to make it easier to overlay (stack or register) the images in post processing. After the images are exposed they are then combined in a graphics program and the complete tonal range of all the properly exposed images are compressed into one image so you see the details in the shadows with out noise and you see the details in the highlights and pure whites with out them being blown out or clipped off. It takes a bit of work and some getting used to the process (compressing tonal range) of all the images but the results can be worth it.
I hope you enjoyed this series on Exposure in Digital Cameras. Check back often or subscribe to the RSS feed for more articles like this on other photography and K10D related subjects.