Module 2: Distances
  Lecture 6: Distance Functions
 

                                            

 

 

Quadratic form distance
  • is a matrix
  • j denotes the similarity of dimension with dimension
  • is positive semi-definite (for distances to be )
 
  • for all non-zero vectors
  • is symmetric (for distances to be symmetric)
  • Example: for color histograms ( is bin-to-bin distance and the maximum)
  • If is an identity matrix, then Euclidean
  • If is a diagonal matrix, then weighted Euclidean
  • Metric?