Bookmark and Share

Computer Vision Metrics: Chapter Four (Part B)

Register or sign in to access the Embedded Vision Academy's free technical training content.

The training materials provided by the Embedded Vision Academy are offered free of charge to everyone. All we ask in return is that you register, and tell us a little about yourself so that we can understand a bit about our audience. As detailed in our Privacy Policy, we will not share your registration information, nor contact you, except with your consent.

Registration is free and takes less than one minute. Click here to register, and get full access to the Embedded Vision Academy's unique technical training content.

If you've already registered, click here to sign in.

See a sample of this page's content below:

For Part A of Chapter Four, please click here.

Bibliography references are set off with brackets, i.e. "[XXX]". For the corresponding bibliography entries, please click here.

Euclidean or Cartesian Distance Metrics

The Euclidean distance metrics include basic Euclidean geometry identities in Cartesian coordinate spaces; in general, these are simple and obvious to use.

Euclidean Distance

This is the simple distance between two points.

Squared Euclidean Distance

This is faster to compute, and omits the square root.

Cosine Distance or Similarity

This is angular distance, or the normalized dot product between two vectors to yield similarity of vector angle; also useful for 3D surface normal and viewpoint matching.

Sum of Absolute Differences (SAD) or L1 Norm

The difference between vector elements is summed and taken as the total distance between the vectors. Note that SAD is equivalent to Manhattan distance.

Sum of Squared Differences (SSD) or L2 Norm

The difference between vector elements is summed and squared and taken as the total distance between the vectors; commonly used in video decoding for motion estimation.

Correlation Distance

This is the correlation difference coefficient between two vectors, similar to cosine distance.

Hellinger Distance

An effective alternative to Euclidean distance, yielding better performance and accuracy for histogram-type distance metrics, as reported in the ROOTSIFT [178] optimization of SIFT. Hellinger distance is defined for L1 normalized histogram vectors as: