Game Development Reference
Considering spatial vision using a function of two-dimensional contrast sensitivity;
Simulation of cortical processes, with a spatial oriented decomposition and a
In fact, this model was introduced to predict the visibility difference between two
images. It was assessed by Myszkowski (1998). It proves to be efficient even though it
does not take into account the vision of colours. The frequency-oriented decomposition
requires a calculation time that cannot be neglected and high memory consumption.
The “Visual Discrimant Model'' by Lubin
This model (Lubin, 1995) tries to simulate the acquisition of the information of
light and the processes applied to the same, from the eyes to the lower level cortical
processing. This model works in four steps:
Simulation of the optic of the eye and the retina by filtering the image and sampling
at 120 pixels per degree;
Calculation of local contrasts by 7 level laplacian decomposition of the image,
then, for each pixel of level i , division by the value of the same pixel of level
Consideration of the orientation (by convolution with four pairs of oriented
filters), followed by calculation of contrast energy;
Transduction phase: Each contrast energy value is standardised by the square of
the contrast detection threshold for the frequency associated with the current level
of the pyramid, and then multiplied by a sigmoid type function (transducer).
This model requires less calibration than the Daly model and its structure is more
logical from the physiological angle. The model is achromatic and decomposition into
28 channels is expensive.
The “Multiscale Model of Adaptation and Spatial Vision'' by Pattanaik et al.
This model (Pattanaik et al., 1998) reproduces all the steps of the course of a light
stimulus inside the human visual system. It has five steps:
Pre-processing of the image: Simulation of the dispersion and diffraction of light
in the ocular globe, followed by obtaining four signals transmitted by cones and
Spatial decomposition of the image by a Difference of Gaussians type of pyramid,
followed by the application of a gain function which increases or decreases the
normalization of the contrast as per the background light (represented by the
value assigned to the pixel corresponding to the upper level of the pyramid);
Transformation of SML channels into AC 1 C 2 channels (A achromatic, C 1 red-
green opposition, blue-yellow opposition C 2 );
Application of a transduction function to consider the contrast sensitivity which
is different for each level of the pyramid;
Addition of signals from rods to channel A .