CS 559: Computer Vision. Lecture 1

Size: px

Start display at page:

Download "CS 559: Computer Vision. Lecture 1"

Melissa Lawson
6 years ago
Views:

1 CS 559: Computer Vision Lecture 1 Prof. Sinisa Todorovic sinisa@eecs.oregonstate.edu 1

2 Outline Gestalt laws for grouping 2

3 Perceptual Grouping -- Gestalt Laws Gestalt laws are summaries of image properties imposed by the constraints of real world Similarity Proximity Closure Good continuation Common fate Figure-ground Symmetry Periodicity 3

4 Gestalt Laws -- Similarity We tend to group objects with similar properties (color, shape, texture) Because objects of the same kind are made of the same matter, and same matter often yields same image properties 4

5 Gestalt Laws -- Proximity We tend to group nearby objects Because matter is cohesive resulting in meaningful configurations of nearby objects 5

6 Gestalt Laws -- Closure We tend to ignore gaps and hallucinate complete, closed contours Because objects occupy finite 3D volumes, and thus have closed surfaces, and so their image projections occupy regions with closed boundaries 6

7 Gestalt Laws -- Good Continuation We prefer to see configurations forming smooth contours Because objects have locally smooth surfaces, and a smooth surface projects in image as a smooth boundary contour 7

8 Gestalt Laws -- Common Fate We tend to see distinct objects with same motion as a unit Because a moving object may consist of parts whose motions depend on the objectʼs motion, and thus project in image coherently 8

9 Gestalt Laws -- Figure-Ground We tend to see certain image areas as foreground or figure and the remaining areas as background or ground Because our knowledge is in terms of distinct objects 9

10 Gestalt Laws -- Symmetry We tend to see symmetric objects as figure against asymmetric background Because many man-made and natural objects are symmetric due to functionality/growth/reproduction processes, and image formation usually preserves symmetry 10

11 Gestalt Laws -- Periodicity We prefer to see objects that periodically repeat Because many man-made and natural objects are spatially periodic due to functionality/growth/reproduction processes, and image formation usually preserves periodicity 11

12 Gestalt Laws -- Periodicity M. sinense bacteria We prefer to see objects that periodically repeat Because many man-made and natural objects are spatially periodic due to functionality/growth/reproduction processes, and image formation usually preserves periodicity 12

13 Which Gestalt Laws Do We Use Here? What Gestalt law should we apply first? Different orderings of the Gestalt laws Different groupings 13

14 From Low-Level Immediately to High-Level Reasoning T-junctions Occlusion 14

15 From Low-Level Immediately to High-Level Reasoning T-junctions Occlusion 14

16 From Low-Level Immediately to High-Level Reasoning T-junctions Occlusion 15

17 From Low-Level Immediately to High-Level Reasoning T-junctions Occlusion 15

18 From Low-Level Immediately to High-Level Reasoning T-junctions Occlusion 15

19 Top-down Interpretation from Context Context and feedback from higher levels resolve low-level ambiguities Because of the context we may not be even aware of alternatives 16

20 Top-down Interpretation from Context Context and feedback from higher levels resolve low-level ambiguities Because of the context we may not be even aware of alternatives 16

21 Top-down Interpretation from Context Context and feedback from higher levels resolve low-level ambiguities Because of the context we may not be even aware of alternatives 16

22 Top-down Interpretation from Context Context and feedback from higher levels resolve low-level ambiguities Because of the context we may not be even aware of alternatives 16

23 Top-down Interpretation from Context Context and feedback from higher levels resolve low-level ambiguities Because of the context we may not be even aware of alternatives 16

Object Perception. 23 August PSY Object & Scene 1

Object Perception. 23 August PSY Object & Scene 1 Object Perception Perceiving an object involves many cognitive processes, including recognition (memory), attention, learning, expertise. The first step is feature extraction, the second is feature grouping