4 Perceiving and Recognizing Objects

Size: px

Start display at page:

Download "4 Perceiving and Recognizing Objects"

Emery Strickland
5 years ago
Views:

1 4 Perceiving and Recognizing Objects Chapter 4 4 Perceiving and Recognizing Objects Finding edges Grouping and texture segmentation Figure Ground assignment Edges, parts, and wholes Object recognition Objects in the brain Development of object perception 4 What Do You See?

2 4 What Do You See? Do you think it is the same house you saw in the previous picture? 4 What Do You See? A more abstract house. Paul Cezanne s Château Noir. 4 Middle Vision How do we organize simple lines and edges into shapes, planes and texture? Edges Low Vision Planes Middle Vision Objects High Vision

4 Finding Edges The computer does a poor job at identifying edges.

3 4 Finding Edges How does the visual system finds edges and puts them together to form shapes, surfaces and planes? 4 Finding Edges The computer does a poor job at identifying edges. 4 Creating Edges Perception of edges when edges are not drawn are called illusory contours.

4 4 Finding Edges Many computer algorithms are available to translate into edges. 4 Finding Edges Sensation and Perception Class Rules for Making Contours Structuralists suggested that atoms of sensation summed up to form perceptual contours.

4 Rules for Making Contours Gestalt psychologists on the other hand suggested that perceptual whole was more than the sum of its sensory parts (atoms).

5 4 Rules for Making Contours Gestalt psychologists on the other hand suggested that perceptual whole was more than the sum of its sensory parts (atoms). And formulated what are called the rules of perceptual groupings (Wertheimer, Köhler, Koffka (1920s 1950s); Palmer and Rock (1990s). 4 Rules for Making Contours Occlusion: The reason why an edge suddenly stops is because an object occludes it. The edge continues under the occluded object. 4 Rules for Linking Contours Good continuation: group elements to form smoothly continuing lines.

4 Perceptual Committees Gestalt Psychologists purposed that the

principles and good guesses to organize perception.

1. Group elements that are similar are perceived one.

6 4 Perceptual Committees Gestalt Psychologists purposed that the brain had neurons that worked as committees that use rules, principles and good guesses to organize perception. 4 Texture Segmentation Linked to two Gestalt grouping principles 1. Group elements that are similar are perceived one. 4 Texture Segmentation 2. Group elements that are proximal (close to each other) perceived together

4 Modern Gestaltism Common region: Elements perceived to be part of a larger region group

7 4 Parallelism/Symmetry Somewhat weaker grouping principles group parallel and symmetric elements together. 4 Modern Gestaltism Common region: Elements perceived to be part of a larger region group together. Connectedness: Elements that are connected to each other group together (Palmer, 1992). 4 Dynamic Grouping Principles Common fate: group elements moving in the same direction together.

4 Dynamic Grouping Principles Synchrony: group

4 Pandemonium Simple model of letter recognition.

8 4 Dynamic Grouping Principles Synchrony: group elements changing at the same time together. 4 Camouflage Which principles of Gestalt are at work here? 4 Pandemonium Simple model of letter recognition. Demons loosely represent neurons; each level represents a different brain area (Selfridge, 1959). A

9 4 Committee Rules: Honor physics and avoid accidents Generally speaking committees come to coherent interpretation of objects in question, however decisions may not be final. Necker s Cube 4 Object Ambiguity Some more examples of ambiguous pictures. 4 More Ambiguity: Accidental Views Any slight shift in accidental viewpoint destroys the perception of the object above.

4 Figure Ground Segmentation What is the

Gestalt figure ground assignment principles:

Meaning and Figure Ground Assignment Is brown or

Object recognition starts before figure ground

10 4 Figure Ground Segmentation What is the to-be-recognized object and background? Gestalt figure ground assignment principles: surroundedness, size, symmetry, parallelism 4 Meaning and Figure Ground Assignment Is brown or white the figure here? Object recognition starts before figure ground assignment finishes 4 Figure and ground: A Complicated Process Which circles are figures and which are holes?

11 4 The Relatability Principle Q: When should we complete edges behind occluders? A: When the edges are relatable by an elbow curve (Kellman and Shipley, 1991) 4 Meaning in the Edges Non-accidental features provide clues to object structure. 4 Is the Whole Seen Before the Parts? Global superiority effect (Navon, 1977). The first goal of the middle vision is to carve large-scale objects and then small scale ones.

4 Parts and Wholes Pairs of concavities indicate part boundaries (Hoffman and

4 Object Recognition: High-Level Vision Processes in object recognition: 1.

Group features into objects ( Middle vision ) 3.

12 4 Parts and Wholes Pairs of concavities indicate part boundaries (Hoffman and Richards, 1984). What happens when cuts can go either way? 4 Object Recognition: High-Level Vision Processes in object recognition: 1. Determine features present in image ( Low-level vision ) 2. Group features into objects ( Middle vision ) 3. Match perceived to encoded representations ( High-level vision ) 4 What Is a Representation? A thing that stands for another thing, for example:

This would be a case of structural description of a representation.

13 4 Naive Template Theory Lock-and-key representations. Templates for objects. Problem: You would need too many templates! 4 Structural Descriptions So a capital A can be defined with three lines. Two forming an angle and the third connecting the two. This would be a case of structural description of a representation. 4 Structural Description Theory Represent the structure of an object, not what it looks like from one view. Recognition-By- Components or geometric icons (RBC or Geons; Biederman, 1987). Take a noodle and put it on to the side of a cylinder and it is a cup; on top it is a pail; put it on a brick and it is an attaché case.

4 The Effect of Viewpoint RBC predicts viewpoint invariance.

4 The Effect of Viewpoint Studies show that as the view point of the

Thus objects may be represented as templates.

14 4 The Effect of Viewpoint RBC predicts viewpoint invariance. Many empirical studies have found viewpoint dependence. 4 The Effect of Viewpoint Studies show that as the view point of the second image is changed it take more time to process. Thus objects may be represented as templates. Greebles 4 Levels of Object Categorization What are these objects? 1. Entry level: birds 2. Subordinate level: sparrow-ostrich 3. Superordinate level: animals

15 4 The Entry Level Which label comes to mind first? The entry level term may be determined by a perceptual committee. 4 Face Recognition What s wrong with this picture? 4 Face Recognition Special processes may be involved in identifying individual faces. Prosopagnosia: Selective inability to recognize faces.

Where/how system: object localization/ manipulation (parietal cortex).

16 4 Objects in the Brain What system: object identification and face recognition neurons (Inferotemporal cortex). Where/how system: object localization/ manipulation (parietal cortex). 4 Grandmother Cells Could a single neuron be responsible for recognizing your grandmother? Animal studies suggest it might be the case.

Occlusion. Atmospheric Perspective. Height in the Field of View. Seeing Depth The Cue Approach. Monocular/Pictorial

Occlusion. Atmospheric Perspective. Height in the Field of View. Seeing Depth The Cue Approach. Monocular/Pictorial Seeing Depth The Cue Approach Occlusion Monocular/Pictorial Cues that are available in the 2D image Height in the Field of View Atmospheric Perspective 1 Linear Perspective Linear Perspective & Texture