Object Recognition

Psy315 - Class 9

(This page last updated 2 October, 2006.)

Visual object recognition is complicated by variations in the retinal image projected by a single object at different times.

• changes in location
• changes in size
• changes in orientation
  • rotation in the picture plane
  • rotation in depth

One theory of visual object recognition: Biederman's Recognition by Components (RBC)

• Start with the assumption that complex shapes are made of combinations of simple shapes.
• Very simple shapes may be recognizable at many different orientations.
• Biederman proposes a set of simple shapes that he calls "geons" to be the basis for object recognition.
• To recognize an object (according to RBC), you identify the geons making it up, along with the spatial relationships among the geons.
  • This is usually enough info to uniquely specify the object.

Other theories of object recognition reject the idea that we break objects down into geons.

They claim that we store separate views of each object in memory.

Face recognition seems to be handled differently from recognition of other objects.

For instance, patients with prosopagnosia can recognize most objects, but not faces.

The Thatcher Illusion (see Fig. 3.20 in textbook) shows that face recognition relies on perceiving the entire configuration as a whole, and that this configural processing is very sensitive to orientation.

 

next class: Attention

Psych 315: Cognitive Psychology Kyle Cave Psychology Dept. U. Mass.