Computational Information Geometry Wonderland

The Earth Mover Distance is one of the key metrics in computer vision and information retrieval based on histograms. The dissimilarity measures dates back from Monge problem. It is a cross-bin measure, that means we do not have to align the bin of the first histogram to the bin of the second histogram distribution.

At ICIP, we shall present some recent work that deals with EMD on a super pixel segmentation tree:

• Takes into account the geometrical and topological structure of segments
• Takes into account unconsistent segmentation from one image to the other
• As fast as solving an EMD problem on 1-D distribution

Here, is the

• poster
• paper

Frank.

Matching feature descriptors in vision is essential for stitching and object recognition among others. Since SIFT is based on discretizing the 360-degree wheel of gradient at different scales, it is better to use circular earth mover distance than a straight EMD.

Experiments are reported Circular Earth Mover?s Distance for the comparison of local features

I write twice this post. When I pushed on the submit button, all my message was erased. This is a big frustration to start a day with such an accident -:)!

Ok, the Earth Mover Distance (EMD) distance introduced in 1997 by Stanford CS group, is in fact known to statisticians under the name of Mallows distance:

It coincides exactly for normalized histograms but not for unormalized distributions. I recommend reading ICCCV'01's paper for a nice description of these similitudes:
The Earth Mover's distance is the Mallows distance: some insights from statistics

Ok, I push the "preview" button and cross fingers for not encountering the same problem twice -:)