A Linear Time Histogram Metric for Improved SIFT Matching

By Ofir Pele, Michael Werman. 2008

372 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

Learning human activities and object affordances from RGB-D videos (2012, methodology): “We also run a SIFT feature based object tracker (Pele and Werman, 2008) to find the corresponding points between the adjacent frames and then compute the transformation matrix based on the matched image points.”
Saliency and Human Fixations: State-of-the-Art and Study of Comparison Metrics (2013, methodology): “She uses a fast implementation of EMD provided by Pele and Werman [20] [21], but without a threshold.”
Wasserstein Distributionally Robust Optimization: Theory and Applications in Machine Learning (2019, background): “Accordingly, the Wasserstein distance is often referred to as the computer vision and graphics [78, 79, 90, 102, 103], data-driven bioengineering [34, 57, 108], clustering [49], dimensionality [18, 35, 87, 93, 95], deep learning with generative adversarial networks [2, 41, 45], domain adaptation…”
The Quadratic-Chi Histogram Distance Family (2010, methodology): “They outperform state of the art distances including χ, QF, L1, L2, ÊMD[14], SIFTDIST[3], EMD-L1[15], Diffusion[16], Bhattacharyya [17], Kullback-Leibler[18] and Jensen-Shannon[19] while having a short running time.”
What Do Different Evaluation Metrics Tell Us About Saliency Models? (2016, methodology): “This has different effects on metrics: for instance, EMD prefers sparser maps provided the predicted locations are near the target locations (the less density to move, the better).”
Fast and robust Earth Mover's Distances (2009, methodology): “Our Rubner’s [29] EMD-L1[20] SIFTDIST[26]”
Faithful Multimodal Explanation for Visual Question Answering (2018, methodology): “We then normalize attention weights over the 14 × 14 resized images to sum to 1, and finally compute the EMD between the normalized attentions and the ground truth.”
Human Activity Learning using Object Affordances from RGB-D Videos (2012, methodology): “We then use SIFT feature matching [25], while enforcing depth consistency across the time frames for obtaining reliable object tracks in the 3D video.”
Unbalanced optimal transport: Dynamic and Kantorovich formulations (2015, background): “It was initially proposed in the computer vision literature to perform image retrieval [27, 21], while its mathematical properties are analyzed in detail by [6, 9].”
A Benchmark of Computational Models of Saliency to Predict Human Fixations (2012, methodology): “We use a fast implementation of EMD provided by Pele and Werman [55] [56] 11 but without a threshold.”
WMDO: Fluency-based Word Mover’s Distance for Machine Translation Evaluation (2019, methodology): “We used the PyEMD wrapper (Mayner, 2019) for calculating the WMD, based on (Pele and Werman, 2008, 2009).”
Combining Quantitative and Logical Data Cleaning (2015, methodology): “Computing the EMD is well-known to be computationally expensive [18, 19].”
Do News Consumers Want Explanations for Personalized News Rankings (2017, background): “There have been further concerns over the effect of such echo chambers [7] on the polarization of society [8, 13].”
Three things everyone should know to improve object retrieval (2012, methodology): “Our combined method sets the new state-of-the-art on all three datasets.”