The Earth Mover's Distance as a Metric for Image Retrieval

By Yossi Rubner, Carlo Tomasi, Leonidas J. Guibas. 2000

5,439 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics Synthesis (2017, methodology): “For this application, the similarity metric d has been chosen to be the Earth Mover’s Distance [43].”
GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering (2019, methodology): “While repeating this operation as we go through the answers, iteratively “moving” probability from the head into the tail [27], we also maintain minimum and maximum ratios between each pair of subsequent answers (sorted by frequency).”
Word Translation Without Parallel Data (2017, background): “We found that this simple criterion is better correlated with the performance on the evaluation tasks than optimal transport distances such as the Wasserstein distance (Rubner et al. (2000)).”
Example Selection For Dictionary Learning (2014, methodology): “ X N more faithfully. To quantify this, we measured the distance of the distribution of selected examples, D(X n), from that of all training examples, D(X N), using the histogram intersection distance[17]. The right columns of Figure 5 D (X n)jjD N)), tends to be lower for ByElement selectors (solid”
A Point Set Generation Network for 3D Object Reconstruction from a Single Image (2016, methodology): “We propose two candidates: Chamfer distance (CD) and Earth Mover’s distance (EMD) [17].”
Wasserstein Adversarial Examples via Projected Sinkhorn Iterations (2019, methodology): “Unlike `p balls, this includes standard image transformations (rotations, translations, distortions), making the Wasserstein distance a more natural metric for images in both monochromatic and RGB image problems (Peleg et al., 1989; Rubner et al., 2000).”
Testing Hypotheses by Regularized Maximum Mean Discrepancy (2013, methodology): “Popular two-sample tests like KolmogorovSmirnov [2] and Cramer-von-Mises [17] are not capable of capturing statistical information of densities with high frequency features.”
Pix3D: Dataset and Methods for Single-Image 3D Shape Modeling (2018, methodology): “Earth Mover’s distance (EMD).”
A Kernel Method for the Two-Sample-Problem (2006, methodology): “ibution in accordance with q, where µis the movement schedule. In general, a large variety of costs of moving mass from xto ycan be used, such as psychooptical similarity measures in image retrieval (Rubner et al., 2000). The following theorem holds (Dudley, 2002, Theorem 11.8.2). Theorem 12 (Kantor”
Visual Causal Feature Learning (2014, background): “Another popular choice has been the Earth Mover’s Distance (Rubner et al., 2000), which measures the distance between two histogram distributions of image statistics.”
MoverScore: Text Generation Evaluating with Contextualized Embeddings and Earth Mover Distance (2019, background): “• We investigate the effectiveness of existing contextualized representations and Earth Mover’s Distance (Rubner et al., 2000) for comparing system predictions and reference texts, leading to our new automated evaluation metric that achieves high correlation with human judgments of text quality.”
The Cramer Distance as a Solution to Biased Wasserstein Gradients (2017, background): “This closeness can matter a great deal: in image modelling, for example, perceptual similarity is key (Rubner et al., 2000; Gao and Kleywegt, 2016).”
GQA: a new dataset for compositional question answering over real-world images (2019, methodology): “While repeating this operation as we go through the answers, iteratively “moving” probability from the head into the tail [33], we also maintain minimum and maximum ratios between each pair of subsequent answers (sorted by frequency).”
Computational Optimal Transport (2018, methodology): “Historically, this corresponds to the original problem posed by Monge in 1781; this setting was also that chosen in early applications of optimal transport in computer vision [Rubner et al., 2000] under the name of “earth mover’s distances”.”
A Survey on Metric Learning for Feature Vectors and Structured Data (2013, background): “. GML (Cuturi & Avis) While χ2-LMNN optimizes a simple bin-to-bin histogram distance, Cuturi and Avis(2011) proposeto consider the morepowerfulcross-bin Earth Mover’s Distance (EMD) introduced by Rubner et al. (2000), which can be seen as the distance between a source histogram x and a destinati”
Large Scale Fine-Grained Categorization and Domain-Specific Transfer Learning (2018, methodology): “3 illustrates calculating the proposed domain similarity by EMD.”
A comparative evaluation of interactive segmentation algorithms (2010, positive): “Once the image has been transformed into an appropriate color space, the next step is to generate a color signature [18] for the known object and background pixels indicated by the user markup.”