Image Database Visualisation and Browsing

Question 1

State three main approaches to visualise image databases.

Answer 1

Graph-based visualisation, Clustering-based visualisation and Mapping-based visualisation.

Question 2

Characterise mapping-based visualisation.

Answer 2

• • High-dimensional CBIR feature vectors
• • Visualise in low-dimensional space
• • Need to get from high-dimensional space to low dimensional space so that original relations are preserved.
• • Dimensionality reduction:
  • • Principal component analysis (PCA)
  • • Multi-dimensional scaling (MDS)

Question 3

Characterise clustering-based visualisation.

Answer 3

• • similar images are grouped together
• • groups can be summarised by representative image(s)
• • content-based
  • –using CBIR features of the images
• -metadata-based
  
  • –uses keywords, filenames etc
• -time-based
  
  • – using time stamp attached to the image upon creation

Question 4

Characterise graph-based visualisation.

Answer 4

• -arranges images in graph structure
  
  • –nodes = images
  • –edges connect images
    
    • –based on shared annotation, image similarity,…
• – visualisation – mass-spring models
  
  • –edges modelled as springs
  • –rest-length based on edge weight/strength
  • –find stable layout (set of differential equations)

Question 5

Describe browsing.

Answer 5

Navigation through visualised image database. It involves interactive and intuitive operations.

Question 6

What are the three main types of browsing?

Answer 6

Horizontal, vertical and time-based