Most input and output devices used in vision research have a non- linear transfer function. That is to say that in the case of input devices the number or voltage produced at the output is not linearly related to the luminance being measured and in the case of output devices, the luminance is not linearly related to the driving voltage or number. Hot-electron tubes such as CRTs are characterised by a transfer function of the form

Luminance = k.(V-V0) ^g

where V is the applied voltage, V 0 is the brightness level and g is about 2.5. For vidicons (TV cameras) g is about 0.5 and for silicon diodes (photocells) g is about 1.

This is an approximate relationship and doesn't apply to other devices such as laser printers and photography where the underlying mechanism of the non-linearity is different. Despite these differences however, they are all considered undesirable as they distort the final stimulus and it is the goal of the vision researcher to remove their influence as far as possible; a process that is known in all cases as gamma correction.

Before a piece of equipment can be gamma corrected the non-linearity must be characterised in such a way that an inverse transfer function can be concocted. The best way to do this will depend on the equipment to hand and how much effort can be spared. If it takes two minutes to perform each measurement one is inclined to devise a system that makes few readings whereas if the system can be fully automated a blunderbuss approach may be more appealing. In either case, sufficient measurements should be taken to ensure that the users error criteria can be met. Gamma correction of CRTs can be performed very adequately by making a dozen or so measurements across the full luminance range and fitting a function of the general gamma form using a standard minimisation technique. For good results the readings should be taken in a dark room with the brightness control on the monitor adjusted such that a zero voltage input gives a just black screen and the contrast control adjusted so that with the maximum voltage the display achieves peak output but doesn't saturate. Once set, these controls should be anchored to prevent the chance of inadvertent adjustment by visitors to the lab. Careful use of a hand-held photometer mounted on a tripod and a general purpose spreadsheet will enable gamma correction within the errors of the original readings. If using a colour display, the three guns should be characterised individually and checked using a white stimulus. If the display is a good quality one the three guns will add together independently.

Figure 9 A typical gamma curve and the signs of mis-adjustment

^ Back to top