Exercise 5: Sensor Linear capture

I must admit to feeling an element of 'interesting, but so what?' When I read through the notes for this exercise. Sensor linear capture refers to the way a digital camera's sensor records an image - basically in camera processing adjusts the tone curve of the image, if it did not do this the image would appear too dark. The exercise requires manipulation of an image's tone curve to simulate this and then the application of the opposite effect on the manipulated image before comparing the results.

Before beginning I decided to do some research in the hope that this might shed some light on why this exercise is important. Martin Evening in "The Adobe Photoshop Lightroom 3 Book" (2010 Adobe Press) provides a good explanation. He asserts that all digital images contain data that is a linear representation of light hitting the sensor:

 "As the light intensity doubles, a sensor records a brightness level that is two times higher...human vision perceives things quite differently from the way a sensor sees and records light...our eyes compensate for increasing levels of brightness in a non linear fashion...Our human vision system is able to adapt to widely changing light levels by constantly compensating for such extremes."1

The way a sensor records light and represents image data looks very dark - even though it may be perfectly exposed. Therefore, in order for a digital capture to look right to our eyes the linear data must at some point be gamma corrected, that is effectively lightened at the mid point.

For further reading on linear capture Evening provides a link to an essay by Bruce Fraser on the Adobe.com website:"Raw capture, linear gamma and exposure."

The first paragraph provides an indication of why linear capture is important, Fraser states that by ignoring the way a digital camera captures an image you run the danger of "failing to exploit the camera's dynamic range, and creating exposures whose shadows are noisier than they need to be." He goes on to explain how levels are distributed in an image using the example of a 12 bit image with 4096 levels and 6 stops of dynamic range:

"If a camera captures six stops of dynamic range, half of the 4,096 levels are devoted to the brightest stop, half of the remainder (1,024 levels) are devoted to the next stop, half of the remainder (512 levels) are devoted to the next stop, and so on. The darkest stop, the extreme shadows, is represented by only 64 levels."

When the cameras processing is applied these levels are redistributed evenly across the image, hence if a photograph is underexposed to maintain the highlight detail the resulting raw conversion will have to open up the shadows and spread the darkest 64 levels over a wider tonal range which runs the risk of exaggerated noise and posterization. He suggests that the best exposure to achieve is one where the highlights are as close as possible to blowing out without actually doing so.

For the exercise I chose an close up image of a lego figure that had a full range of evenly distributed tones  in the histogram:

No adjustments

Linear image - curve adjusted

For this image I opened the photograph in Photoshop, converted to 16bit, made a curves adjustment with a steep but smooth downward curve which resulted in an under exposed image which simulates sensor linear capture.

'Restored' image - curve adjusted

For this image, I opened the saved version of the simulated linear capture image again converted to 16bit and made a curve adjustment opposite to that applied in the previous image, that is pushed up to increase brightness in a smooth curve.

I compared this image with the original until I achieved an equivalent result with a similar histogram.

No adjustment detail

Restored curve detail

Finally I examined the original and restored images at high magnification on screen. I found that there was evidence of increased noise and posterization in the shadow areas of the restored image. (The detail shots above show this a little although the effect is not pronounced.) I concluded that by performing these extreme adjustments on this image I was simulating trying to increase exposure on an underexposed image. I have always held the view, picked up from where I do not know, that it is important in digital photography to preserve highlight detail above all else as that cannot be recovered in post processing. I always believed that an underexposed image was preferable to an overexposed picture as underexposure is easier to compensate for in software. I guess this remains true, however, I am surprised at how few tones in an image are allocated to the dark and midtones. Correct exposure it seems is more of a minefield than I first thought the key to which I believe is understanding the image you want to create at the time of capture, understanding which tonal values are most important to the pictures success and using knowledge and experience to achieve this. In truth I rarely worry about exposure, my camera usually does a pretty good job and I always shoot raw so can alter the exposure in post processing if need be. This exercise is beginning to make me understand how little I actually consider exposure at the time of pressing the shutter and that I need to learn more about how my camera behaves if I want to produce optimum results.

At the end of this exercise I feel I understand more of how a digital camera's sensor captures an image. The true importance of this however is something I am guessing will become more significant as I explore the next few exercises which are concerned with highlight clipping, noise and dynamic range.

1 p.629 The Adobe Photoshop Lightroom 3 Book: The Complete Guide for Photographers by Martin Evening (2010 Adobe Press)