visual object perception

• 70-90% of brain is used for visual object perception
• The retinal image is not sufficient to specify what's in the environment. It's in 2 dimensions, doesn't give you depth or shade, has countless possible interpretations
• Vision is a process that produces from images of the external world a description that is useful to the viewer and not cluttered with irrelevant information. In the case of human vision, the initial representation is in no doubt- it consists of arrays of image intensity values as detected by the photoreceptors in the retina: 'Marr, 1976'
• We are edge detectors


  • We detect edges on small receptive fields and combine them into larger receptive in order to comprehend the image
  • We take segments and build them into a whole
• The 2D images are ambiguous due to shading, occlusion, viewpoint, lightness, darkness. Edges can be unclear, so we need to make inferences by detecting features, grouping parts into similar objects, parsing scene so that figures can be identified from background, identifying object

  
• Steps to solution


  1. Get info

  2. Group features into wholes

  3. Select foreground of scene from the background
• Gestalt grouping principles


  Rules for organising features into coherent objects. Can be thought of as perceptual heuristics based on what the most likely outcome is from immense amount of perceptual experience. The perceptual system has learned that these principles predict the correct answer most of the time
• The whole is different from the sum of its parts. In perception, we need to see the whole rather than the parts, so we need to explain how parts (sensations) become wholes - perceptual organisation
• Law of Prägnaz (good figure)


  Every stimulus pattern is seen in such a way that the resulting structure is as simple as possible. We prefer to see things in the environment as simple, regular, symmetrical and close
• Law of transposition


  We process the entire figure, not the part. Parts are interchangeable
• Laws of grouping


  • Similarity - objects that are similar are grouped together
  • Good continuation - we group lines together to create lines that create the smoothest, straightest path
  • Proximity - objects that are close together are grouped together
  • Common fate - objects that move together are grouped together
  • Closure - we assume that shapes are regular and symmetrical, we close shapes to make them into wholes
• Figure ground


  Which edge belongs where - our interpretation of a scene depends on foreground and background, not constant, can change with attention. Objects lower in the display more likely to be seen as figure. Once figure ground determined, the figure is more thing-like and memorable than the ground, the figure is in front of the ground, the ground is seen as uniform and extends beyond the picture, the contour separating figure from the ground belongs to the figure
• If we see a pattern that looks like a face, we'll automatically see a face because social info is so important to us
• We process faces as a whole, not the sum of its parts.