Factors affecting brightness and colour vision under water

Abstract

Both theoretical and practical importance can be attached to attempts to model human threshold and supra-threshold visual performance under water. Previously, emphasis has been given to the integration of visual data from experiments conducted in air with data of the physical specification of the underwater light field. However, too few underwater studies have been undertaken for the validity of this approach to be assessed. The present research therefore was concerned with the acquisition of such data. Four experiments were carried out: (a) to compare the predicted and obtained detection thresholds of achromatic targets, (b) to measure the relative recognition thresholds of coloured targets, (c) to compare the predicted and obtained supra-threshold appearance of coloured targets at various viewing distances and under different experimental instructions, (d) to compare the predicted and obtained detection thresholds for achromatic targets under realistic search conditions. Within each experiment, observers were tested on visual tasks in the field and in laboratory simulations. Physical specifications of targets and backgrounds were determined by photometry and spectroradiometry. The data confirmed that: (a) erroneous predictions of the detection threshold could occur when the contributions of absorption and scattering to the attenuation of light were not differentiated, (b) the successful replication of previous findings for the relative recognition thresholds of colours depended on the brightness of the targets, (c) the perceived change in target colour with increasing viewing distance was less than that measured physically, implying the presence of a colour constancy mechanism other than chromatic adaptation and simultaneous colour contrast; the degree of colour constancy also varied with the type of target and experimental instructions, (d) the successful prediction of the effects of target-observer motion and target location uncertainty required more than simple numerical corrections to the basic detection threshold model. It was concluded that further progress in underwater visibility modelling is possible provided that the tendency to oversimplify human visual performance is suppressed.