The latest developments in display technology allowed the extension of the range of luminance levels that can be reproduced on the screen. With the popularisation of displays offering controllable display illumination, modern screens can now show lumi- nance levels that are both higher and lower than before. The same display can now show images spanning the photopic (daylight), mesopic (dusky) and scotopic (dark) luminance levels at the same time. This introduces new challenges for the currently existing models of human vision, which are mostly fitted to the data collected in bright, photopic conditions. These models cannot predict the changes in colour and contrast perception that happen at low luminance. The human visual system loses sensitivity in the dark, with contours losing sharpness and colours becoming washed out. Only a model based on the physiological differences between day and night vision can accurately predict these changes. In this thesis, a novel algorithm is introduced that can predict how the human perception changes with absolute luminance. The model is based on previous psy- chophysical measurements as well as our own experiments. Our measurements were conducted using a high dynamic range display that was calibrated as a part of this re- search. The predictions of the model were verified in subjective evaluation experiments to ensure their accuracy. The proposed model has two major applications. In the first application, the pre- dicted contrast and colour changes at low luminance can be introduced to an image viewed at high, photopic luminance to simulate the appearance of night scenes on a display that otherwise would not be able to invoke the dark perception phenomena. In the second application, the predicted appearance changes can be compensated for so that a processed image shown on a dimmed display viewed in dark surroundings can be perceived as closely as possible to what the original image would look like on a bright display. The results show that the proposed algorithm performs equally well or better than similar algorithms for all static images and video clips used in our comparisons. These results indicate that the underlying model of perceived changes caused by night vision can predict such changes accurately.