Presentation is loading. Please wait.

Presentation is loading. Please wait.

The Earth Mover's Distance

Similar presentations


Presentation on theme: "The Earth Mover's Distance"— Presentation transcript:

1 The Earth Mover's Distance
9/17/2018 The Earth Mover’s Distance as a Metric for Image Retrieval and as a Metric for Distributions with Applications to Image Databases Ranjini Swaminathan University of Arizona 17 September September 2018

2 The Problem To employ EMD as a method to find dissimilarities between images. Image Parameters : Images are described by features such as color (3-D representation) Texture (a distribution of energy in the frequency domain) 17 September September 2018

3 Need for methods Image Retrieval-based on dissimilarity between two images. Steps: - Compute histograms/signatures - Implement EMD /other method - Navigation 17 September September 2018

4 Histograms A histogram {hi} is a mapping from a set of d-dimensional integer vectors i to the set of nonnegative reals. An example: - Consider a grey-level histogram where d=1 - split possible grey values into N intervals - image pixels fall into respective bins 17 September September 2018

5 Histograms contd. Disadvantage:
-fail to strike a balance between expressiveness and efficiency. Adaptive Binning of Histograms: -Need prior knowledge of distribution of features for all images. -significant information restricted to few bins. 17 September September 2018

6 Signatures Signature {sj=(mj,wj) } - a set of clusters
mj - d-dimensional mode Wj - number of pixels belonging to the cluster A histogram is a special case of a signature. 17 September September 2018

7 Signatures contd. Clusters 17 September September 2018

8 Methods Bin-By-Bin Dissimilarity measures - Minkowski-Form Distance:
dLr(H,K) = (hi-kir)1/r - Histogram Intersection: d(H,K) = 1 - i min(hi,ki)/i ki 17 September September 2018

9 Methods contd. - K-L Divergence and Jeffrey Divergence:
dKL(H,K)= i hi log hi/ki dJ(H,K)= i (hi log hi/mi+ ki log ki/mi) - X2 Statistics: d2(H,K) = i(hi-mi)2/mi 17 September September 2018

10 Methods contd. Cross-Bin Dissimilarity measures
Quadratic-form distance Match distance - Kolmogorov-Smirnov distance 17 September September 2018

11 Signatures vs Histograms
- fixed size structures - efficiency vs expressiveness Signatures - representative of features - less info yet better retrieval 17 September September 2018

12 EMD -measures the amount of work done in moving mass over holes in the same space. -not a matching problem but a transportation problem -increased efficiency 17 September September 2018

13 The Earth Mover's Distance
9/17/2018 Flow equations WORK(P,Q,F)= ijdijfij with the constraints fij  0 j fij  wpi i fij  wqj ijfij = min(i wpi , j wqj ) where 1  i  m , 1  j  n 17 September September 2018

14 EMD-Features EMD(P,Q) = ijdijfij/ijfij Features:
-applies to signatures which subsume histograms - avoids quantization problems -allows partial matches & variable length descriptions -metric 17 September September 2018

15 MDS Multi Dimensional Scaling as a Perceptual Evaluation Tool:
ij is the original distance between objects i and j dij is the distance between the points in a low dimensional space STRESS = (i,j(dij - ij )2/ i,j ij 2)1/2 Zero Stress =perfect fit 17 September September 2018

16 Color Courtesy : 17 September September 2018

17 Color Distributions Given :Set of images Do : cluster
obtain signatures compute EMD with a stress value Result : images aligned along lightness and chroma MDS displays more subtle similarities between closely related images 17 September September 2018

18 Color distributions contd.
Test results show that EMD with signatures outperforms all other methods. Navigation : - Embeddings are adaptive - Zooming into query - Use of “don’t cares” 17 September September 2018

19 Efficiency EMD efficiency depends on: Transportation algorithm
Easy to compute lower bounds However, EMD is : Computationally intensive Works better for smaller sample sizes 17 September September 2018

20 Notes Metric : (Merriam-Webster) Proof :
“a mathematical function that associates with each pair of elements of a set a real nonnegative number with the general properties of distance such that the number is zero only if the two elements are identical, the number is the same regardless of the order in which the two elements are taken, and the number associated with one pair of elements plus that associated with one member of the pair and a third element is equal to or greater than the number associated with the other member of the pair and the third element.” Proof : Triangle Inequality 17 September September 2018

21 Future Work Other vision problems –classification, recognition and segmentation Outside computer vision???? -in speech recognition -in biology 17 September September 2018


Download ppt "The Earth Mover's Distance"

Similar presentations


Ads by Google