Shapes and Objects
Vocabulary:
Patient D.F.: Got carbon monoxide poisoning due to an improper functioning hot water heater, resulting in visual changes and preserved cognitive abilities. Could see colors and textures, but couldn’t discriminate between shapes, though if asked to reach for objects she would reach with fingers spaced the proper length. Could properly perceive shape in terms of motor capacity but not cognition. Demonstrates how there are two visual pathways, an intact dorsal pathway and impaired ventral stream
Patient R.V.: Had damage to the dorsal stream, could recognize objects but not reach for them, intact ventral stream
Ventral stream: A visual pathway for object recognition, gets more input from the parvocellular pathway, possibly encodes information about local image features
Dorsal pathway: A visual pathway for action, gets more input from the magnocellular pathway, potentially responsible for global shapes
Double dissociation: Describes how damage to either the dorsal or ventral stream only impacts one kind of function (action/recognition)
Lateral connections: Connect the dorsal and ventral visual streams physically, allow for sharing of information between the two
Gestalt psychology: Describes that the whole of something is different from the individual parts that make it up (sum of its parts), the general idea is that we organize a visual scene into objects by using physical and spatial relationships between elements to determine what goes together. Argues that familiarity with an object wasn’t relevant
Committee system: A system of thinking and visual perception that takes in various aspects, cues need to be integrated together but one doesn’t necessarily dominate the other. Each cue is a ‘committee member’ and ‘presents a case’ towards the visual system, and ‘considers evidence’ from each cues strength (if 3 suggest one thing and 1 another, the three cues are chosen. Also if there are two cues for different interpretations but 2 are stronger, that interpretation is chosen). Committee leader is convexity and VP is surroundedness, members are familiarity and symmetry, homogeneity shows up once in a blue moon
Figure: The main object in a given scene or visual space, will ‘own’ borders if they are physically in front of them, knowing this allows for depth perception and definition of figure shape
External contour: The boundary between the figure and ground, the border is ‘owned’ by whatever is physically forward
Ground: The area behind an object in a given scene or visual space, will border the figure but may not own a border if they are behind the object, background/environment
Occlude: When something covers something else, one figure may cover the ground, and own boundaries
Convexity: Whether an object is concave/convex, the region is better seen as a figure if it is convex as opposed to concave, more convex regions means more likely to be figure. Most important cue for figure/ground assignment
Homogenous: Similar, like for color. If a region is the same color it is more likely ground, if it is different colors it is the figure
Heterogenous: Different, like in terms of color
Symmetry: Can aid in figure/ground distinguishment, the more sides match the more likely it is the figure. Many objects are symmetrical (like living creatures) which is why we see symmetrical things as more likely to be figures. Less important of a cue than convexity
Surroundedness: Whether one region encloses another, cue for figure/ground where surrounding region is the ground and interior is the figure
Familiarity: Argued that it might aid in figure/ground distinguishment, the more someone sees something the more likely they are to recognize it, originally argued irrelevant by Gestalt psychologists though data shows familiarity is important (rotational studies)
Amodal completion: Happens when an occluder is “seen” as covering an object, happens when the occluder is the same color as the background, creates illusory contour
Identity hypothesis: Though they are seen as different, the underlying computations for many forms of contour interpolation are the same, demonstrated by the Petter effect. Lengths are first determined, then determined as illusory or occludent. States that the same rules/computations are used for contour interpolation across different kinds of shapes. Evidenced by illusory and occlued contours showing edges when together
Contour interpolation: How the brain “sees” a shape across a gap, continues lines, operates regardless of color/texture
Petter effect: Demonstrates that the smaller an illusory contour is, the more likely it is as been seen
Junction: The point between surfaces, where one surface (with a certain color/texture) ends and another begins, tends to be evidence of occlusion, edge intersection, tend to be in L or Y shape
Surface interpolation: Interpolation of things like color/texture behind an occluder, depends on oriented edges. Similar surface properties means various objects are more likely to be seen as one (color), and they are in close proximity
Abstract: The inability of shape to be described as the collection of all its parts, can be more or less than its sum of parts
Constant curvature theory: Shapes are broken up into segments of similarity, and encoded based on their similarity to one another rather than individual parts
Factors for seeing something as a cohesive object for a Gestalt Psychologist, not sufficient on their own but as a whole make up recognition:
Similarity: How related are elements/fragments to each other, includes things like shape, color, and texture, size
Proximity: How close objects are to one other in comparison to closeness to other objects
Common fate: More relevant when objects are in motion, are they moving in the same direction or different/static
Good continuation: Looks at object edges and orientations, how things change over time, are relative positions and patterns contiguous if inferred to continue behind an occluded
How the brain makes shapes (contour interpolation):
Identify junctions (where a contour stops/started)
Assume interpolated edges begin at junctions
Decide if there is sufficient reliability between junction contours to satisfy smoothness constraints
Draw straight lines (even if the edge is curved) from edge 1 (E1) and edge 2 (E2) towards each other in their given directions
If they intersect, and the angle is greater than or 90 degrees. Intersection must be will the extensions, not the actual line.
The visual system represents less information than what’s available in the visual field about shape, along with invariance to change in size, position, and orientation. It can also reason about how the shape could change due to movement of parts (person walking will still be one object but change boundaries)
The visual system focuses on relations between elements instead of the elements themselves, ignoring small contour variators to focus on the bigger/more important ones. It describes how shape primitives are spatially related (curvature)