Comparing Top<i>k</i>Lists

Fagin, Ronald; Kumar, Ravi; Sivakumar, D.

doi:10.1137/s0895480102412856

Cited by 671 publications

(637 citation statements)

References 6 publications

Supporting

Mentioning

616

Contrasting

Unclassified

Order By: Relevance

“…Using an existing human-derived higher-resolution similarity matrix [10] we were able to fully rank the results which in turn allowed us to assess the ability of features to estimate perceived similarity more thoroughly. (Note for this comparison we used the "G" measure [23] that takes into account rank order). We also applied the PMIF feature set to two more popular tasks: sketch-based image retrieval (SBIR) (a) (b) Fig.…”

Section: Discussionmentioning

confidence: 99%

“…For the retrieval based assessment we compared the rankings of the computational and human-derived retrievals (which excluded the query image) using the 7 measure (7 ∈ [0, 1]) [23].…”

Section: Experimental Designmentioning

confidence: 99%

“…We did this for the top ) ∈ {10,20,40,603 retrieved textures. The 7 measure has the advantage that it considers the relative rankings within the two retrievals compared with traditional measures: precision and recall [23] which do not.…”

Section: Experimental Designmentioning

confidence: 99%

“…In this framework, human ranking data are used as the ground-truth and Kendall's rank correlation coefficients [24] are used as the performance measures. We also employed the 7 measure [23] used in the texture retrieval task as it is suitable for comparing two non-identical rankings [23]. As in [21], we set _ = 5 and low and high thresholds for the Canny detector [7] to 0.05 and 0.2 respectively.…”

Section: A Sketch-based Image Retrieval Experimentsmentioning

confidence: 99%

See 3 more Smart Citations

Perceptually Motivated Image Features Using Contours

Dong

Chantler

2016

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

Abstract-Dong et al. examined the ability of 51 computational feature sets to estimate human perceptual texture similarity, however, none performed well for this task. While it is well-known that the human visual system is extremely adept at exploiting longer-range aperiodic (and periodic) "contour" characteristics in images, none of the investigated feature sets exploit higher order statistics (HOS) over larger image regions (>19×19 pixels). We therefore hypothesise that long-range HOS, in the form of contour data, are useful for perceptual texture similarity estimation.We present the results of a psychophysical experiment that shows that contour data are more important, than local image patches, or global 2nd-order data, to human observers for this task.Inspired by this finding, we propose a set of perceptually motivated image features (PMIF) that encode the long-range HOS computed from spatial and angular distributions of contour segments. We use two perceptual texture similarity estimation tasks to compare PMIF against the 51 feature sets referred to above and four commonly used contour representations. This new feature set is also examined in the context of two additional tasks: sketch-based image retrieval and natural scene recognition. The results show that the proposed feature set performs better, or at least comparably to, all the other feature sets. We attribute this promising performance to the fact that the proposed feature set exploits both short-range and long-range HOS.

show abstract

Section: Discussionmentioning

confidence: 99%

“…For the retrieval based assessment we compared the rankings of the computational and human-derived retrievals (which excluded the query image) using the 7 measure (7 ∈ [0, 1]) [23].…”

Section: Experimental Designmentioning

confidence: 99%

Section: Experimental Designmentioning

confidence: 99%

Section: A Sketch-based Image Retrieval Experimentsmentioning

confidence: 99%

See 2 more Smart Citations

Perceptually Motivated Image Features Using Contours

Dong

Chantler

2016

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

show abstract

“…Let ℝ + be the set of all positive real numbers and ℝ * be the set of extended real numbers (Definition 2.1 page 4 [Blumenthal(1953)] pages 12-13, [Laos(1998)] pages 118-119 36 This is the method of "inscribed polygons" for calculating the length of a curve and goes back to Archimedes: [Brunschwig et al(2003) [Sherstnev(1962)], page 4, [Schweizer and Sklar(1983)] page 9 ⟨(1.6.1)-(1.6.4)⟩, [Bessenyei and Pales(2014) [Kirk and Shahzad(2014)] page 113 ⟨Definition 12.1⟩, [Deza and Deza(2014)] page 7, [Hoehn and Niven(1985)] page 151, [Gibbons et al(1977)Gibbons, Olkin, and Sobel] page 51 ⟨square-meanroot (SMR) (2.4.1)⟩, [Euclid(circa 300BC)] ⟨triangle inequality-Book I Proposition 20⟩ 39 metric space: [Dieudonné(1969)], page 28, [Copson(1968)], page 21, [Hausdorff(1937)] page 109, [Fréchet(1928)], [Fréchet(1906)] page 30 near metric space: [Czerwik(1993)] page 5 ⟨b-metric; (1),(2),(5)⟩, [Fagin et al(2003a)Fagin, Kumar, and Sivakumar], [Fagin et al(2003b) …”

Section: Examplesmentioning

confidence: 99%

Properties of distance spaces with power triangle inequalities

Greenhoe

2016

Preprint

View full text Add to dashboard Cite

Metric spaces provide a framework for analysis and have several very useful properties.Many of these properties follow in part from the triangle inequality. However, there are several applications in which the triangle inequality does not hold but in which we may still like to perform analysis. Properties of distance spaces with power triangle inequalities DANIEL J. GREENHOEAbstract: Metric spaces provide a framework for analysis and have several very useful properties. Many of these properties follow in part from the triangle inequality. However, there are several applications in which the triangle inequality does not hold but in which we may still like to perform analysis. This paper investigates what happens if the triangle inequality is removed all together, leaving what is called a distance space, and also what happens if the triangle inequality is replaced with a much more general two parameter relation, which is herein called the "power triangle inequality". The power triangle inequality represents an uncountably large class of inequalities, and includes the triangle inequality, relaxed triangle inequality, and inframetric inequality as special cases. The power triangle inequality is defined in terms of a function that is herein called the power triangle function. The power triangle function is itself a power mean, and as such is continuous and monotone with respect to its exponential parameter, and also includes the operations of maximum, minimum, mean square, arithmetic mean, geometric mean, and harmonic mean as special cases.2010 Mathematics Subject Classification 54E25 (primary); 54A05,54A20 (secondary)

show abstract

Bibliography

2016

Handbook of Bibliometric Indicators

View full text Add to dashboard Cite

Comparing TopkLists

Cited by 671 publications

References 6 publications

Perceptually Motivated Image Features Using Contours

Perceptually Motivated Image Features Using Contours

Properties of distance spaces with power triangle inequalities

Bibliography

Contact Info

Product

Resources

About