Grouping

• Goals
  • Gather features that belong together
  • obtain an intermediate representation that cmopactly describes key image or video parts
• Top down segmentation: pixels belong together because they are from the same object
• Bottom up segmentation: pixels belong together because they look similar
• Gestalt Theory
  • Whole is greater than sum of its parts
  • Relationships among parts can yield new properties/features

Segmentation

• K-means clustering
• Probabilistic clustering
• Mixture of Gaussians
• GMM
• Mean-shift Algorithm