Theories of Optical Illusions

Optical illusions have long been studied by scholars of perception. The idea is that we can in find out how correct perception works, by observing situations in which misperception occurs. Visual perception is complex. It includes a large set of processes that start at the formation of images on our eyes and provide us with an interpretation of the objects and events in the scene. Roughly, we distinguish three kinds of processes: image formation; building geometric models of the spatio-temporal scene; and interpreting the scene in terms of the concepts we know. According to this division, we classify illusions of vision into physiological illusions due to the image formation or very early representation of images, optical illusions in the classic sense, which are about perceiving surfaces and motions different from the real objects, and cognitive illusions resulting from an ambiguous interpretation.

My work has been on optical illusions in the classic sense, which are due to the processes of making models of the scene, or as computational scientists say, the processes of “reconstructing the scene.” Illusions in this category include the geometric-optic illusions, illusions of lightness and color, visual completion illusions, and illusions of motion.

Working on of signal and scene reconstruction, I have found there are certain limitations to the underlying computations. Perfect signal reconstruction is not always possible. This gave rise to a few theories of illusions.

Theory of Bias

The images and videos recorded on our retina are not perfect (like any other signal recorded by some device). Using the language of statistics, we can say the signal is noisy. The computations of our vision system then can be considered as estimations and analyzed with statistics. It turns out that in the  presence of noise the estimations are statistically biased, causing simple image features, image  motion, and 3d shape to be estimated away from the veridical.

Read more at this Website

Causal Filtering

A class of patterns, composed of repeating patches of asymmetric intensity profile, elicit strong perception of illusory motion. Note, the illusory effect is perceived in the periphery of the eye, not near the center of fixation. It depends on the eye's spatial resolution – the human eye has higher resolution at the fovea than the periphery. We propose that the main cause of this illusion is erroneous estimation of image motion induced by fixational eye movements. Image motion is estimated with energy filters, which are symmetric in space, but asymmetric (causal) in time. That is, only the past, but not the future, is used to estimate the temporal energy. We have shown that such filters mis-estimate the motion in patterns like the one shown here (which are locally asymmtric) at certain spatial frequencies.

More in this Paper

Interplay of 3D Motion Estimation and Segmentation

The pattern shown is a variant of Enigma (by the painter Leviant), and it causes perception of illusory movement in the bands. Our explanation relies on a model for the interpretation of three-dimensional motion information contained in retinal motion measurements. Eye movements and/or accommodation changes cause weak retinal motion signals, which are interpreted by higher-level processes in a way that gives rise to Enigma and a larger class of patterns exhibiting the same illusory movement, which we created. These patterns are such that their texture does not allow recovery of all the components of the observer's own motion.

More in this Paper