A generalized metric distance between hierarchically partitioned images

Manouvrier, Maude; Rukoz, Marta; Jomier, Geneviève

doi:10.1145/1133890.1133894

Cited by 6 publications

(5 citation statements)

References 25 publications

(100 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar concerns have already emerged in the field of Image Retrieval, where surveys can be found in [21,23]. In addition, to increase the relevance of our results, we believed it was logical and judicious to develop a suitable subdivision of the fields (as suggested by [15,13]). So we spatially subdivided the daily fields into patches corresponding to known precipitation drivers.…”

Section: Discussionsupporting

confidence: 55%

Using an expert deviation carrying the knowledge of climate data in usual clustering algorithms

Biabiany¹,

Page²,

Bernard³

et al. 2020

Preprint

View full text Add to dashboard Cite

In order to help physicists to expand their knowledge of the climate in the Lesser Antilles, we aim to identify the spatio-temporal configurations using clustering analysis on wind speed and cumulative rainfall datasets. But we show that using the L2 norm in conventional clustering methods as K-Means (KMS) and Hierarchical Agglomerative Clustering (HAC) can induce undesirable effects. So, we propose to replace Euclidean distance (L2) by a dissimilarity measure named Expert Deviation (ED). Based on the symmetrized Kullback-Leibler divergence, the ED integrates the properties of the observed physical parameters and climate knowledge. This measure helps comparing histograms of four patches, corresponding to geographical zones, that are influenced by atmospheric structures. The combined evaluation of the internal homogeneity and the separation of the clusters obtained using ED and L2 was performed. The results, which are compared using the silhouette index, show five clusters with high indexes. For the two available datasets one can see that, unlike KMS-L2, KMS-ED discriminates the daily situations favorably, giving more physical meaning to the clusters discovered by the algorithm. The effect of patches is observed in the spatial analysis of representative elements for KMS-ED. The ED is able to produce different configurations which makes the usual atmospheric structures clearly identifiable. Atmospheric physicists can interpret the locations of the impact of each cluster on a specific zone according to atmospheric structures. KMS-L2 does not lead to such an interpretability, because the situations represented are spatially quite smooth. This climatological study illustrates the advantage of using ED as a new approach.

show abstract

Section: Discussionsupporting

confidence: 55%

Using an expert deviation carrying the knowledge of climate data in usual clustering algorithms

Biabiany¹,

Page²,

Bernard³

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…In Manouvrier et al (2005), it has been presented for images represented by full-balanced integer, x ∈ {0, 1, 2, 3, 4, 5, 6, 7, 8} N set of all nodes (resp. segments) identifiers N ( ) set of all nodes (resp.…”

Section: Distances Between Multi-level Feature Vectorsmentioning

confidence: 99%

“…It can be progressively approximated by comparing both sub-trees level by level, following a breadth first order. This characteristic is very useful to filter images using different levels of precision Lin et al, 2001;Lu et al, 1994;Wang Z et al, 2003) -see in Manouvrier et al (2005) for more details about multi-level filtering. ( ) (i, j) is defined as the -distance between multi-level feature vectors i and j, where for all nodes n (n ∈ N ) appearing from root level (level 0) to a level , w n > 0, and where for all nodes n appearing in a deeper level ( > ), w n = 0.…”

Section: ( ) -Distance: Approximatingmentioning

confidence: 99%

“…The root node stores the global feature vector of the entire image. A multi-level feature vector could be used as a multi-level filtering structure or a coarse-to-fine representation Lin et al, 2001;Lu et al, 1994;Manouvrier et al, 2005): each image is compared with the query image using their global signature and if they are similar enough (according to a given threshold and to a similarity metric), then the corresponding image segments are compared, and so on, level by level. For example, using such multi-level signatures with images of Fig.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Δ-distance: A family of dissimilarity metrics between images represented by multi-level feature vectors

2006

Self Cite

View full text Add to dashboard Cite

This article presents the -distance, a family of distances between images recursively decomposed into segments and represented by multi-level feature vectors. Such a structure is a quad, a quin or a nona-tree resulting from a fixed and arbitrary image partition or from an image segmentation process. It handles positional information of image features (e.g. color, texture or shape). -distance is the generalized form of dissimilarity measures between multi-level feature vectors. Using different weights on tree nodes and different distances between nodes, distances between trees or visual similarity between images can be computed based on the general definition of . In this article, we present three -based distance families: two families of distances between tree structures, called T -distance (T for Tree) and S-distance (S for Segment), and a family of visual distances between images, called V-distance (V for Visual). The V-distance visually compares two images using their tree representation and the other two distances compare the tree structures resulting from image segmentation. Moreover, we show how existing distances between multi-level feature vectors appear to be particular cases of the -distance.Keywords Image database . Distance between quad/quin or nona-trees . Similarity of images . Similarity of image segments . Content-based image retrieval

show abstract

“…In the image processing area we have several works by Yu et al (2006) which propose a robust method for distances metrics for estimation of similarity. Manouvrier et al (2005) propose a distance between images recursively partitioned into structures of trees, which is effective in CBIR techniques. Georgiou et al (2007) propose a metric distance between distributions in the image segmentation process.…”

Section: Introductionmentioning

confidence: 99%