Professional Documents
Culture Documents
April-01-11
- The red exclamation marks denote assumptions that most likely need empirical justification (experiment)
7:36 PM - The light bulbs denote assumptions that might need empirical justification but not necessarily
- The question marks denote questions that I believe are somewhat important and should possibly be
considered for the experiment
- The star denotes certain ideas that might be superfluous but might be helpful
So what's going on!?
f1
- As soon as the vertical disc pair move inwards and occlude the horizontal
sides of the rectangle, illusory contours begin to appear (f2)
It is likely that discs alone would not suffice for contours and that it
is necessary to have the vertical dot pair move outwards in order for
the contours to be perceived
Because without the small circles, the visual system can easily
infer that the occlusion is taking place because the grey circles
are blocking it from the observer but the appearance of the
small circles brings in a new element that might lead to the
perception of the contours
- Also ---> as soon as the grey dot pairs (horizontal and vertical) leave the
rectangle, the discs occlude the sides and this occluding is simultaneous
with the grey dots leaving the rectangle
The visual system might attribute the occlusion to the smaller dots f2
instead of the discs ---> why?
- Probably due to the cues given by the small circles (to be
shortly discussed)
○ But when only the vertical discs occlude the sides of the
rectangle, then the visual system uses the grey dots to establish
the boundaries of the contours
□ Evidence that the small circles are necessary
------------------------------------------------------------------------------------------------------
○ P(CO|I) = P(I|CO)*P(CO)
○ P(~CO|I) = P (I|~CO)*P(~CO)
Input
- There has to be something about the input that increases the
probability of there being a camouflaged object
○ Cues in the demo
The Heuristics
A priori in the sense that they tell their effects are in some sense pre -
established before any sort of experience and in that sense they are
related to the prior probability factor (priori probability factor)
A posteriori in the sense that the cues are available through experience
and vary depending on the input processed by the visual system < --
which depends on the observer's experience (likelihood factor)
- History
○ Gestalt school (late nineteen to early twentieth century)
"the perceptual whole is more than the sum of all its sensory parts"
- Occlusion
○ Seems to play a big role in causing the illusory contours
○ When the edges of an object or part of the object are suddenly cut off, the visual system
keeps a representation of the object as a whole ---> that is, the visual system interprets
the sudden disappearance of edges in an image projected onto the retinae as if another
object has obstructed the light reflecting from it ---> this heuristic has a very well-defined
adaptive function and that is to allow observers in the world to continue to have a holistic
representation of objects even though the images projected onto the retinae are seldom
whole - any given scene projected onto the retinae usually contains a multitude of objects
which usually have many parts blocked by other objects depending on their relative
distance to the observer.
The visual system's use of this cue is evident in Kanizsa figures which are images that
take advantage of the heuristic and lead observers to perceive illusory contours:
□ The image reflected onto the retinae contains certain shapes that are
systematically cut off in order to portray the sense that another shape (usually
a simple square or triangle) is blocking the cut off sections and obstructing the
reflection of light from those parts. An innate understanding of the physics of
the world regarding the opacity of solid objects is perhaps influencing the
perception of the observers. (pg87) Certain top-down mechanisms seem to be
causing the observer to perceive contours representing the outline of the
inferred occluding shape, but these contours are purely illusory because there
○ What is interesting about the demo is that the objects that occlude the rectangular outline
are clearly represented as grey discs. That is, the need for the visual system to create
contours in order to represent an image occluding the rectangle does not seem to be
required since the grey discs are clearly responsible. Illusory contours are needed for the
visual system to make sense of occluded figures but why are they needed even when the
occluding figures are prominently displayed?
The four small circles are activating another heuristic that is responsible for the
contours
- The visual system computes that it is more probable for there to be a third object
(represented by the illusory contours), given the occlusion done by the four discs
and the movement of the four small circles.
- Why does the visual system makes this conclusion?
Speculations:
Perhaps the asynchrony of the two pairs of discs has something to
do with it ---> maybe it is more improbable that something in the
world would have this kind of movement than if there is another
object involved
Related to the implicit understanding of the physics of the
world?
- Common fate
Explanation:
• when parts of a design are all moving in the same direction look like they are all part of
the same unit
• in a design or layout, the directional lines that move in the same direction are said to
have a common fate or destiny
• when two objects are in the same direction in a layout, the directional lines become
dominant within the design
• these directional lines can point a viewer’s gaze in a particular direction
- The parts that move together are then grouped together and perceived as being a single
object
○ We propose that the visual system's inference about objects in the 3D world (based on the
2D image), is an automatic decision making process which can be represented by the
Bayesian probability principle.
- The probability that the illusory contours are caused by a single object is likely due
to the heuristics that are feeding the likelihood factor which group the four small
circles together
Not a whole lot must be said about these but they are also worth mentioning
- Symmetry
○ Both sets can be said to create symmetrical patterns
- Similarity
○ Equal similarity for grey discs and small circles
- Law of proximity
○ Objects closer to each other are more likely to be grouped
together than items that are more widely separated
Grey discs are more distant from each other in
comparison with the small circles
- Law of familiarity
○ Very speculative but it's possible that the motion of the small
circles seem somewhat familiar to the visual system
Especially in relation to the motion of the grey discs
Their motion seems more biological
- Figure-ground --- "is a region A in front of a region B" (pg88)
○ Size
Small circles are much smaller than the grey discs AND
the symmetrical shape that the four combined make is
much smaller than a shape that connecting the grey
discs would make
○ Surroundedness
The grey discs can be said to contain the small circles
○ These might have to do with the "biological" motion of the four small circles
Hypothesis:
As represented by the Bayesian theorem, we hypothesize that the visual system works in a
probabilistic manner and certain heuristics alter the likelihood factor in the equation and invoke the
perception of an illusory figure ; the combination of occlusion and common fate in conjunction with
the similarity and proximity cue are responsible for the visual system's grouping of the four small
circles and the attribution of occlusion to their illusory contours.