Two types of signals (top-down and bottom-up) are assumed in this model. Top-down signals are hypothesized to be derived from a higher level of the visual system and represent the predictive visual information. Bottom-up signals are hypothesized to be derived from a lower level of the visual system and represent the original sensory information. Subtraction occurs between the two signals; the resultant prediction error (delta) is input to a higher level of the visual system that operates the top-down signal to minimize the prediction error.
Motion Aftereffect can also be well explained by the subtraction algorithm in Delta Model, because the subtraction can induce the inverted motion direction observed in Motion Aftereffect. The same logic can be applied to various kinds of Aftereffect illusions.
Please refer to the following paper for detail.
Watanabe, E., Matsunaga, W., and Kitaoka, A., Motion signals deflect relative positions of moving objects, Vision Research 50, 2381-2390 (2010)
The delta model is conceptually equivalent to the predictive coding theory (below) for the cerebral visual system. The model is the general extension of the idea of predictive coding to the entire brain. This theory will explain the cause of many illusions. In the above-mentioned 2010 paper, we are trying to explain illusions by delta model or predictive coding. This attempt will be one of the earliest descriptions about relationship between illusions and predictive coding.
Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects
Rajesh P. N. Rao & Dana H. Ballard
Nature Neuroscience volume 2, pages 79–87 (1999)