Part 1: Summary of a Euclidean Distance Paper

Liwei Wang, Yan Zhang, Jufu Feng. On the Euclidean Distance of Images (TMI, August 2005)
IMage Euclidean Distance (IMED): Robust to small perturbations
Terminology in the paper is very similar to ours
Similar figures that show degradation and measures of distance
Context is Eigenfaces and face recognition using the Face Recognition Technology (FERET) database and digit classification
Compete with tangent distance and Hausdorff distance
Fast (simple to compute), maybe obeys the triangle inequality
Efficiently embedded in powerful image recognition techniques (SVM, LDA, PCA, etc.)
Standardizing Transform (ST) - transform domain smoothing
Smoothing noiseless images can increase recognition rate
The method is said to be efficient, but no evidence is included

Roy Schestowitz 平成17年6月21日