Forensics - Maximising Image Content

Scope and Objective

In this posting I am exploring the most effective way to manipulate an image to maximise image content for identification purposes.

Tonal Range and Histograms

The tonal range of an image is simply the range of shades in the image from dark to bright.

While 256 tones (or levels) may be the maximum tonal range of an 8-bit digital image, there is no guarantee that we will be able to see all of these tones, if indeed they are even needed.  For a start, the gamut of the display device, coupled with the brightness and contrast settings may curtail the available tonal range.  Then we have to consider the ambient lighting where the image is being observed.  People may be reading this posting on a tablet or on a smart phone.  Strong ambient lighting may overwhelm the tonal range of the screen.  For best results I recommend reading this post and viewing the images at a desk, away from distracting lighting, such as daylight coming through a window.

More importantly, the content of an image will determine it's tonal range.  In nature, pure black and pure white are rarely encountered, yet they form part of the standard tonal range.  On a dull, foggy day images will tend to be flat-looking, with perhaps fewer than 100 discrete tones required to create the full image.  Most if not all of these tones will fall in the midtone range.  By contrast, on a sunny, summer's day the dynamic range (the gap between the darkest and brightest tones) of the ambient light will far surpass the recording capabilities of a standard camera sensor, and therefore it will be impossible to capture the full range of tones within a single image.  This manifests as high contrast images with detail clipped, perhaps often at both ends of the histogram.

An image histogram, for those unfamiliar with it, is simply a graphical representation of an image based on it's tonal content, looking specifically at the quantity of pixels occurring at each tonal level.  If an image has been well exposed the image data will fall roughly in the middle of the histogram without touching the left or right hand edges.  The shape of the histogram is dictated by image content - there is no ideal-looking histogram.

Poorer quality images very often have issues which can be diagnosed simply by studying the histogram.  For instance in the image below this Soft-plumaged Petrel Pterodroma mollis appears over exposed in the original JPEG image from the camera.  The histogram confirms the worst - the highlights fall off the edge of the chart and are 'clipped' from the image.  During JPEG compression clipped information tends to be discarded to help shrink the image file size.  This data is irretrievable.

While use of the levels tool can darken the image and retrieve some detail (lower left image), the histogram confirms that highlights remain clipped.  Not only has encoded data been lost at the edge of the tonal range, data which existed throughout the tonal range in the original RAW file, but which wasn't expressed in the original JPEG, is likely to be discarded.  Data loss in the midtone range tends to become obvious when image contrast is increased.  Data gaps appear in the histogram as spikes, referred to as a 'combing effect' (see lower left hand image histogram below).

Luckily I was shooting in RAW as well as JPEG when I photographed this bird.  By creating a new JPEG manually using Camera RAW software I have heen able to retrieve highlight data together with a lot more tonal information not presented in the original JPEG.  For more on working in RAW and a video demonstration of Camera Raw workflow see HERE.

In summary, the solution to maximising image detail for ID purposes is simple - shoot and process the image in RAW.

Soft-plumaged Petrel Pterodroma mollis, off Argentina.