PDF version of this document

Measuring Inter-image distances

Euclidean Distance:

The simple method for comparing disagreement among two images is to take their per-pixel difference. This is sensitive to small misalignments though.

Shuffle Distance:

In our experiments we have defined image distance as the shuffle distance between two images, as illustrated below

3#3 = 17#17 The computation of the shuffle distance

2#2

Shown below are the effects of the shuffle transform:

3#3 = 18#18 The product of a shuffle transform

3#3 = 19#19 The effect of increasing the shuffle radius