Digitized Images: What Type of Grey Scale Should One Use?

Abstract
One of the issues faced by engineers when designing a system which records an external event and represents it in the form of a digitized image on a VDU screen is which type of grey scale to use. An experiment is described which compares, in a simulated digitized image, the effect of a linear and a logarithmic grey scale on the detectability of a straight-line signal embedded in visual noise. It was found that both bright and dark signals were detected more easily with the linear scale. A signal detection theory analysis was carried out to compare human performance with that of an ‘ideal’ observer who performed the detection task with a filter spatially matched to the signal. It was found that the model of performance for this ideal observer accounted well for the results provided the assumption of a linear transformation of luminance was made. The analysis showed that the superiority of the linear over the logarithmic grey scale was simply due to the higher signal-to-noise ratio of the signals in the former.