The Prague Texture Segmentation Datagenerator and Benchmark - Algorithms
optimized
for

The concise description of algorithms contains hyperlinks to further information (author, algorithms details, BIB entry, WWW external page).
The algorithm features are: f1 = classification (supervised segmentation), f2 = hiearchy result (manual selection), f3 = known number of regions, f4 = [reserved].
xaos's
GMRF+EM
version 2.0
BIB
WEB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
M. Haindl & S. Mikes:   Gaussian Markov random field model

An efficient and robust type of unsupervised multispectral texture segmentation method is presented. Single decorrelated monospectral texture factors are assumed to be represented by a set of local Gaussian Markov random field (GMRF) models evaluated for each pixel centered image window and for each spectral band. The segmentation algorithm based on the underlying Gaussian mixture (GM) model operates in the decorrelated GMRF parametric space. The algorithm starts with an oversegmented initial estimation which is adaptively modified until the optimal number of homogeneous texture segments is reached.
xaos's
AR3D+EM
version 1.0
BIB
WEB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
M. Haindl & S. Mikes:   3D autoregressive random field model

A new unsupervised multispectral texture segmentation method with unknown number of classes is presented. Multispectral texture mosaics are locally represented by four causal multispectral random field models recursively evaluated for each pixel. The segmentation algorithm is based on the underlying Gaussian mixture model and starts with an over segmented initial estimation which is adaptively modified until the optimal number of homogeneous texture segments is reached. The performance of the presented method is extensively tested on the Prague segmentation benchmark using the commonest segmentation criteria and compares favourably with several alternative texture segmentation methods.
test's
Blobworld
BIB
WEB
f1: 0
f2: 0
f3: 0
f4: 0
Ch. Carson & M. Thomas & S. Belongie & J. M. Hellerstein & J. Malik:   Blobworld: A System for Region-based Image Indexing and Retrieval

Blobworld is a system for content-based image retrieval. By automatically segmenting each image into regions which roughly correspond to objects or parts of objects, we allow users to query for photographs based on the objects they contain.
test's
EDISON
version 1.1
BIB
WEB
f1: 0
f2: 0
f3: 0
f4: 0
Ch.M. Christoudias & B. Georgescu & P. Meer:   Edge Detection and Image SegmentatiON (EDISON) System

This system is a low-level feature extraction tool that integrates confidence based edge detection and mean shift based image segmentation. It was developed by the Robust Image Understanding Laboratory at Rutgers University.
test's
JSEG
BIB
WEB
f1: 0
f2: 0
f3: 0
f4: 0
Y. Deng & B.S. Manjunath:   Unsupervised Segmentation of Color-Texture Regions in Images and Video

A method for unsupervised segmentation of color-texture regions in images and video is presented. This method, which we refer to as JSEG, consists of two independent steps: color quantization and spatial segmentation. In the first step, colors in the image are quantized to several representative classes that can be used to differentiate regions in the image. The image pixels are then replaced by their corresponding color class labels, thus forming a class-map of the image. The focus of this work is on spatial segmentation, where a criterion for ?good? segmentation using the class-map is proposed. Applying the criterion to local windows in the class-map results in the J-image, in which high and low values correspond to possible boundaries and interiors of color-texture regions. A region growing method is then used to segment the image based on the multiscale J-images. A similar approach is applied to video sequences. An additional region tracking scheme is embedded into the region growing process to achieve consistent segmentation and tracking results, even for scenes with nonrigid object motion. Experiments show the robustness of the JSEG algorithm on real images and video.
scarpa's
TFR
BIB
WEB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
G. Scarpa & M. Haindl:   Texture Fragmentation and Reconstruction

The Texture Fragmentation and Reconstruction (TFR) segmentation algorithm is based on a texture modeling particularly suited for segmentation in an unsupervised framework. A texture is regarded for each fixed spatial direction as a finite-state Markov chain where the states of the process are quantized colors. On the basis of this modeling, a simple segmentation algorithm is derived that precesses independently color and spatial information, by first performing a color-based clustering, which provides the quantized colors, and then by means of a further spatial-based clustering, which separates regions according to their transition probability profile. Finally, a region merging algorithm allows to recover the different textures, that is to recompose their internal Markov chains.
scarpa's
TFR/KLD
BIB
WEB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
G. Scarpa & M. Haindl & J. Zerubia:   A Hierarchical Finite-State Model for Texture Segmentation

It is an improved version of the TFR algorithm where the region gain has been changed by introducing a Kullback-Leibler Divergence (KLD) term modeling the region similarity in terms of spatial location.
xaos's
AR3D+EM multi
version 1.0
BIB
WEB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
M. Haindl & S. Mikes:   Multi 3D autoregressive random field model

A novel unsupervised multi-spectral multiple-segmenter texture segmentation method with unknown number of classes is presented. The unsupervised segmenter is based on a combination of several unsupervised segmentation results, each in different resolution, using the sum rule. Multi-spectral texture mosaics are locally represented by four causal multi-spectral random field models recursively evaluated for each pixel. The single-resolution segmentation part of the algorithm is based on the underlying Gaussian mixture model and starts with an over segmented initial estimation which is adaptively modified until the optimal number of homogeneous texture segments is reached. The performance of the presented method is extensively tested on the Prague segmentation benchmark using the commonest segmentation criteria and compares favourably with several alternative texture segmentation methods.
test's
EGBIS
BIB
WEB
SRC
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Pedro F. Felzenszwalb and Daniel P. Huttenlocher:   Efficient Graph-Based Image Segmentation

We define a predicate for measuring the evidence for a boundary between two regions using a graph-based representation of the image. We then develop an efficient segmentation algorithm based on this predicate, and show that although this algorithm makes greedy decisions it produces segmentations that satisfy global properties. We apply the algorithm to image segmentation using two different kinds of local neighborhoods in constructing the graph, and illustrate the results with both real and synthetic images. The algorithm runs in time nearly linear in the number of graph edges and is also fast in practice. An important characteristic of the method is its ability to preserve detail in low-variability image regions while ignoring detail in high-variability regions.
felipecalderero's
GSRM sup.
version BHAT/KL area-weighted/-unweighted
BIB
WEB
DOC
f1: 0
f2: 0
f3: 1
f4: 0
Felipe Calderero:   General statistical region merging - supervised - 10 bins

General region merging technique based on a size-weighted/-unweighted direct statistical measure of the empirical distributions of the regions, using the Kullback-Leibler divergence/Bhattacharyya coefficient.

This version is supervised, meaning that the number of regions for the evaluated partitions was manually set to the number of regions in the ground truth partitions.

In this implementation, empirical distributions were quantized to 10 bins.
felipecalderero's
GSRM unsup.
version BHAT/KL area-weighted/-unweighted
BIB
WEB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Felipe Calderero:   General statistical region merging - unsupervised - 10 bins

General region merging technique based on a size-weighted/size-unweighted direct statistical measure of the empirical distributions of the regions, using the Kullback-Leibler divergence/Bhattacharyya coefficient.

This version is UNSUPERVISED, meaning that the number of regions is automatically selected using a significance index.

In this implementation, empirical distributions were quantized to 10 bins.
test's
SWA
version def_par
BIB
WEB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
E. Sharon & M. Galun & D. Sharon & R. Basri & A. Brandt:   SWA algorithm

SWA algorithm segmentation by weighted aggregation, is derived from algebraic multigrid solvers for physical systems, and consists of fine-to-coarse pixel aggregation. Aggregates of various sizes, which may or may not overlap, are revealed as salient, without predetermining their number or scale.
test's
HGS
version E/W/C
BIB
WEB
f1: 0
f2: 0
f3: 0
f4: 0
Minh A. Hoang & Jan-Mark Geusebroek & Arnold W.M. Smeulders:   HGS algorithm

The HGS unsupervised segmenter is based on the integration of the Gabor filters with the measurement of color. Single versions of the method differ in their photometric invariance power (HGS-E no invariance, HGS-W low, HGS-C full invariance). The spatial frequency is measured by sampling the incoming image with a shifted Gaussian in the spatial frequency domain, and the color is measured by sampling the signal with Gaussian in wavelength domain. The method implies that the color-texture is measured in the wavelength-Fourier domain. The measurement filter in this domain boils down to a 3D Gaussian, representing a Gabor-Gaussian in the spatial-color domain.
xaos's
MW3AR
BIB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
M. Haindl, S. Mikeš, P. Pudil:   Hierarchy 3D autoregressive random field model

An unsupervised multi-spectral, multi-resolution, multiple-segmenter for textured images with unknown number of classes is presented. The segmenter is based on a weighted combination of several unsupervised segmentation results, each in different resolution, using the modified sum rule. Multi-spectral textured image mosaics are locally represented by four causal directional multi-spectral random field models recursively evaluated for each pixel. The single-resolution segmentation part of the algorithm is based on the underlying Gaussian mixture model and starts with an over segmented initial estimation which is adaptively modified until the optimal number of homogeneous texture segments is reached.

sylvia's
TEX-ROI-SEG
version 1.0
BIB
WEB
f1: 0
f2: 0
f3: 0
f4: 0
Michael Donoser and Horst Bischof:   Texture ROI-Segmentation

Defaut Parametrization

[1] Donoser, M. and Bischof, H. (2008). Using Covariance Matrices for Unsupervised Texture Segmentation. In Proceedings of International Conference on Pattern Recognition (ICPR) , Tampa, USA.

[2] Donoser, M. and Bischof, H. (2007). ROI-SEG: Unsupervised color segmentation by combining differently focused sub results. In Proceedings of Conference on Computer Vision and Pattern Recognition (CVPR), Minneapolis, USA.

Implementation by Sylwia Steginska
xaos's
AR3D+EM ii
BIB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
M. Haindl & S. Mikes & P. Vacha:   Illumination Invariant Unsupervised Segmenter

A novel illumination invariant unsupervised multispectral texture segmentation method with unknown number of classes is presented. Multispectral texture mosaics are locally represented by illumination invariants derived from four directional causal multispectral Markovian models recursively evaluated for each pixel. Resulted parametric space is segmented using a Gaussian mixture model based unsupervised segmenter. The segmentation algorithm starts with an over segmented initial estimation which is adaptively modified until the optimal number of homogeneous texture segments is reached. The performance of the presented method is extensively tested on the large illumination invariant benchmark from the Prague Segmentation Benchmark using 21 segmentation criteria and compares favourably with an alternative segmentation method.
sv.pons's
ac-multi
version 1.0
BIB
SRC
DOC
f1: 0
f2: 0
f3: 1
f4: 0
Sandro Vega-Pons; José L. Rodriguez:   Active contour based multiclass texture image segmentation algorithm

jlgil's
Tx-acM
version vers 2
BIB
f1: 0
f2: 0
f3: 1
f4: 0
Vega-Pons, S.; Gil-Rodríguez, J.L. and Vera-Pérez, O.L.:   Active Contour Algorithm for Texture Segmentation

Tx-acM is an algorithm for unsupervised texture segmentation based on the active contour without edges model with level set representation and a connected component filtering strategy to noise reduce inside each functional minimization step. A set of texture features calculated using several texture models: first order statistic model, co-occurrence matrix model, run length matrix model, Gabor's descriptors (1D and 2D) and moment's descriptors are incorporated discreetly in a vector of valued images at the input of the algorithm. See reference [1] for details.
The prefix "Tx" makes reference to "Tx Estudio", a system oriented to textured image segmentation making use of the paradigms: tone, texture and/or tone+texture. Several non-supervised image segmentation algorithms are implemented into the system: Tx-acM, Tx-acB, Tx-kMeans, Tx-fuzzykMeans,Tx-MeanShift, Tx-ART2, Tx-fuzzyART and Tx-SOM.

[1] Vega-Pons, S.; Gil-Rodríguez, J.L and Vera-Pérez, O.L. (2008). "Active contour algorithm for texture segmentation using a texture feature set". In 19th International Conference on Pattern Recognition. ICPR2008. IEEE Computer Society. TuBCT8.32, (1-4), 2008. ISSN: 1051-4651, ISBN: 978-1-4244-2174-9. DOI:10.1109/ICPR.2008.4761583
scarpa's
R-TFR/M
version 1
BIB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
R. Gaetano, G. Scarpa, G. Poggi:   Recursive TFR (with manual selection)

The Texture Fragmentation and Reconstruction (TFR) algorithm, recently proposed for the segmentation of textured images, has been applied with promising results to high-resolution remote-sensing images. The algorithm provides a sequence of nested segmentation maps which allow the analysis at various scales of observation. However, the performance which is very good at large scales, with complex semantic areas retrieved with remarkable accuracy, becomes less satisfactory at finer scales. By using the TFR in a recursive fashion, segmenting the image in just two regions, initially, with each region further segmented only if relevant subregions emerge, we get the Recursive TFR (R-TFR). R-TFR allows one to better adapt to local statistics and to extract significant textures also at finer scales. In this version the best segmentation scale is manually selected.
chaththa85's
ImprvGMRF
version 2
BIB
DOC
f1: 0
f2: 0
f3: 1
f4: 0
C. Dharmagunawardhana, S. Mahmoodi, M. Niranjan, M. Bennett:   Gaussian Markov random field based improved texture descriptor for image segmentation

An improved semi parametric method of texture feature formulation using GMRFs.

Texture descriptor based on Gaussian Markov random fields (GMRFs). A spatially localized parameter estimation technique using local linear regression is performed and the distributions of local parameter estimates are constructed to formulate the texture features. The inconsistencies arising in localized parameter estimation are addressed by applying generalized inverse, regularization and an estimation window size selection criterion. The texture descriptors are named as local parameter histograms (LPHs) and are used in texture segmentation with the k-means clustering algorithm.

The segmentation results on general texture datasets demonstrate that LPH descriptors significantly improve the performance of classical GMRF features and achieve better results compared to the state-of-the-art texture descriptors based on local feature distributions.
scarpa's
R-TFR/K
version 1
BIB
DOC
f1: 0
f2: 0
f3: 1
f4: 0
R. Gaetano, G. Scarpa, G. Poggi:   R-TFR (with known number of classes)

The Texture Fragmentation and Reconstruction (TFR) algorithm, recently proposed for the segmentation of textured images, has been applied with promising results to high-resolution remote-sensing images. The algorithm provides a sequence of nested segmentation maps which allow the analysis at various scales of observation. However, the performance which is very good at large scales, with complex semantic areas retrieved with remarkable accuracy, becomes less satisfactory at finer scales. By using the TFR in a recursive fashion, segmenting the image in just two regions, initially, with each region further segmented only if relevant subregions emerge, we get the Recursive TFR (R-TFR). R-TFR allows one to better adapt to local statistics and to extract significant textures also at finer scales. In this version the number of classes is known a priory.
scarpa's
TS-MRF/M
version 1
BIB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
C. D'Elia, G. Poggi, G.Scarpa:   Tree-structured Markov Random Field (with manual selection)

The Tree-Structured Markov Random Field (TS-MRF) algorithm is a spectral-based (hence not texture-based) classifier which utilizes MRF prior models to get regularized segmentations.
The image is recursively segmented in smaller and smaller regions until a stopping condition, local to each region, is met. Each elementary binary segmentation is obtained as the solution of a MAP estimation problem, with the region prior modeled as an MRF. Since only binary fields are used, and thanks to the tree structure, the algorithm is quite fast, and allows one to address the cluster validation problem in a seamless way. In addition, all field parameters are estimated locally, allowing for some spatial adaptivity.
In this version the proper segmentation scale (tree pruning) is left to the user.
scarpa's
TS-MRF/K
version 1
BIB
DOC
f1: 0
f2: 0
f3: 1
f4: 0
C. D'Elia, G. Poggi, G.Scarpa:   Tree-structured Markov Random Field (with known number of classes)

The Tree-Structured Markov Random Field (TS-MRF) algorithm is a spectral-based (hence not texture-based) classifier which utilizes MRF prior models to get regularized segmentations.
The image is recursively segmented in smaller and smaller regions until a stopping condition, local to each region, is met. Each elementary binary segmentation is obtained as the solution of a MAP estimation problem, with the region prior modeled as an MRF. Since only binary fields are used, and thanks to the tree structure, the algorithm is quite fast, and allows one to address the cluster validation problem in a seamless way. In addition, all field parameters are estimated locally, allowing for some spatial adaptivity.
In this version the number of classes is an input parameter.
scarpa's
DHC/M
version 1
BIB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
G. Scarpa, G. Masi, R. Gaetano, L. Verdoliva, G. Poggi:   Dynamic Hierarchical Classifier (with manual selection)

Recursive tree-structured segmentation is a powerful tool to deal with the non-stationary nature of images. By fitting model parameters to each region/class under analysis one can adapt the segmentation algorithm to the local image statistics, thus improving accuracy. However, a single model/segmenter cannot fit regions with wildly different nature, and one should be allowed to select in a suitable library the tool most suited to the local statistics. The dynamic segmentation/classification algorithm (DHC), uses two segmenters, based on spectral and textural properties, respectively, and a suitable rule for switching model locally.
In this version the segmentation is hand-picked from the hierarchical segmentation stack.
scarpa's
DHC/K
version 1
BIB
DOC
f1: 0
f2: 0
f3: 1
f4: 0
G. Scarpa, G. Masi, R. Gaetano, L. Verdoliva, G. Poggi:   Dynamic Hierarchical Classifier (with known number of classes)

Recursive tree-structured segmentation is a powerful tool to deal with the non-stationary nature of images. By fitting model parameters to each region/class under analysis one can adapt the segmentation algorithm to the local image statistics, thus improving accuracy. However, a single model/segmenter cannot fit regions with wildly different nature, and one should be allowed to select in a suitable library the tool most suited to the local statistics. The dynamic segmentation/classification algorithm (DHC), uses two segmenters, based on spectral and textural properties, respectively, and a suitable rule for switching model locally.
In this version the number of classes is given as input.
cpanag@csd.uoc.gr's
Results_vote_Class_merge
WEB
f1: 0
f2: 0
f3: 0
f4: 0
Costas Panagiotakis, Ilias Grinias and Georgios Tziritas:   Texture Segmentation Based on Voting of Blocks, Bayesian Flooding and Region Merging

We propose an unsupervised texture image segmentation framework with unknown number of regions, which involves feature extraction and classification in feature space, followed by flooding and merging in spatial domain. The distribution of the features for the different classes are obtained by a block-wise unsupervised voting framework using the blocks grid graph or its minimum spanning tree and the Mallows distance. The final clustering is obtained by using the k-centroids algorithm. An efficient flooding algorithm is used, namely, Priority Multi-Class Flooding Algorithm (PMCFA), that assign pixels to labels using Bayesian dissimilarity criteria. Finally, a region merging method, which incorporates boundary information, is introduced for obtaining the final segmentation map. The proposed scheme is executed for several number of regions, we select the number of regions that minimize a criterion that takes into account the average likelihood per pixel of the classification map and penalizes the complexity of the regions boundaries. Segmentation results on the Prague benchmark data set demonstrate the high performance of the proposed scheme.
jolen217's
deep_brain_model
version 1.0
BIB
WEB
f1: 0
f2: 0
f3: 0
f4: 0
Nan Zhao:   Deep Brain Model

Deep brain model is an unsupervised segmentation framework with unknown number of classes simulating the deep structure of the primate visual cortex. This model is based on a deep scale space in which a pool of receptive field models in pre-cortical processing and early vision is applied in each scale to produce feature maps. The graph-based image segmentation is then employed to select object boundaries among the edges of superpixels.
hnizdja2's
LevelSet
version 1,0
BIB
SRC
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Jan Hnízdil, Šimon Hlaváč, Tomáš Kuzin:   MI-ROZ LevelSetSegmentation

School work on FIT ČVUT.
Implementation of pattern recognition algorithm based on IEEE article:

Level Set Segmentation With Multiple Regions
written by Thomas Brox and Joachim Weickert
perutond's
mean-shift
version 1.0
BIB
SRC
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Ondřej Perutka, David Vaníček:   mean-shift

Image segmentation based on the mean-shift algorithm.
bartejak's
Energy Minimization with Label Costs
version 1.0
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Jakub Bartel, Libor Jonát, Pavel Zbytovský:   Energy Minimization with Label Costs

Implementace algoritmu podle článku Fast Approximate Energy Minimization with Label
Costs
xaos's
AR3D+EM dyn
BIB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Haindl Michal, Mikeš Stanislav:   Unsupervised Dynamic Textures Segmentation

An unsupervised dynamic colour texture segmentation method uses unknown and variable number of texture classes. Single regions with dynamic textures can furthermore dynamically change their location as well as their shape. Individual dynamic multispectral texture mosaic frames are locally represented by Markovian features derived from four directional multispectral Markovian models recursively evaluated for each pixel site. Estimated frame-based Markovian parametric spaces are segmented using an unsupervised segmenter derived from the Gaussian mixture model data representation which exploits contextual information from previous video frames segmentation history. The segmentation algorithm for every frame starts with an over segmented initial estimation which is adaptively modified until the optimal number of homogeneous texture segments is reached.
scarpa's
eCog
BIB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
Martin Baatz and Arno Schäpe:   eCognition

From abstract:
The approach [...] aims for an universal high-quality solution applicable and adaptable to many problems and data types. As each image analysis problem deals with structures of a certain spatial scale, the average image objects size must be free adaptable to the scale of interest.
This is achieved by a general segmentation algorithm based on homogeneity definitions in combination with local and global optimization techniques. A scale parameter is used to control the average image object size. [...]
scarpa's
ENVI
BIB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
Jin Xiaoying:   ENVI

A digital image can be processed by an image processing method that calculates a gradient map for the digital image, calculates a density function for the gradient map, calculates a modified gradient map using the gradient map, the density function and the selected scale level, and segments the modified gradient map. Prior to segmenting the modified gradient map, a sub-image of the digital image can be segmented at the selected scale level to determine if the selected scale level will give the desired segmentation.
sabattom's
NCuts
version 1.0.0
SRC
DOC
f1: 0
f2: 0
f3: 1
f4: 0
Bc. Krákora Vojtěch, Bc. Šabata Tomáš:   Normalized Cuts

Normalized Cuts and Image Segmentation
Jianbo Shi and Jitendra Malik, Member, IEEE
richtto6's
Histogram ratio features
version 1.0
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Tomas Richtr:   Histogram ratio features for color texture classification

Implementation of classification by article Histogram ratio features for color texture classification for MI-ROZ.
stameser's
Autocorrelation function
version 0.9
SRC
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Petrou:   Autocorrelation function

Auto-correlation function looks in image as on random field and calculates how it's correlated with itself.
More details in book (see link below). Page 196.
frydatom's
OBLRCD
version 1
SRC
DOC
f1: 0
f2: 0
f3: 1
f4: 0
Tomáš Frýda:   Object boundary location by region and contour deformation

Object boundary location by region and contour deformation

Postprocessing
nenenevg's
LBP
version 13
WEB
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Ojala et al. 1996:   Local binary patterns

Local binary patterns (LBP) is a type of feature used for classification in computer vision. LBP is the particular case of the Texture Spectrum model proposed in 1990. Local Binary Pattern (LBP) is a simple yet very efficient texture operator which labels the pixels of an image by thresholding the neighborhood of each pixel and considers the result as a binary number. Due to its discriminative power and computational simplicity, LBP texture operator has become a popular approach in various applications. It can be seen as a unifying approach to the traditionally divergent statistical and structural models of texture analysis.
xiaofang's
LocalGlobalGraph
version color
BIB
DOC
f1: 0
f2: 1
f3: 0
f4: 0
Xiaofang Wang, Yuxing Tang, Simon Masnou, and Liming Chen:   A Global/Local Affinity Graph for Image Segmentation

Construction of a reliable graph capturing perceptual grouping cues of an image is fundamental for graph-cut based image segmentation methods. We propose a novel sparse global/local affinity graph over superpixels of an input image to capture both short and long range grouping cues, thereby enabling perceptual grouping laws, e.g., proximity, similarity, continuity, to enter in action through a suitable graph cut algorithm. Moreover, we also evaluate three major visual features, namely color, texture and shape, for their effectiveness in perceptual segmentation and propose a simple graph fusion scheme to implement some recent findings from psychophysics which suggest combining these visual features with different emphases for perceptual grouping.

Specifically, an input image is first oversegmented into superpixels at different scales. We postulate a gravitation law based on empirical observations and divide superpixels adaptively into small, medium and large sized sets. Global grouping is achieved using medium sized superpixels through a sparse representation of superpixels' features by solving a `0-minimization problem, thereby enabling continuity or propagation of local smoothness over long range connections. Small and large sized superpixels are then used to achieve local smoothness through an adjacent graph in a given feature space, thus implementing perceptual laws, e.g., similarity and proximity. Finally, a bipartite graph is also introduced to enable propagation of grouping cues between superpixels of different scales.
cbampis's
GRPNMF
BIB
SRC
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Christos G. Bampis & Petros Maragos & Alan C. Bovik:   Projective non-negative matrix factorization for unsupervised graph clustering

Unsupervised graph clustering and image segmentation algorithm based on non-negative matrix factorization. It considers arbitrarily represented visual signals (in 2D or 3D) and uses a graph embedding approach for image or point cloud segmentation. It extends a Projective Non-negative Matrix Factorization variant to include local spatial relationships over the image graph. By using properly defined region features, this method can be applied for object and image segmentation.
xaos's
MW3AR8^i
BIB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Michal Haindl, Stanislav Mikes, and Mineichi Kudo:   Unsupervised Surface Reflectance Field Multi-Segmenter

An unsupervised, illumination invariant, multi-spectral, multi-resolution, multiple-segmenter for textured images with unknown number of classes is presented. The segmenter is based on a weighted combination of several unsupervised segmentation results, each in different resolution, using the modified sum rule. Multi-spectral textured image mosaics are locally represented by eight causal directional multi-spectral random field models recursively evaluated for each pixel. The single-resolution segmentation part of the algorithm is based on the underlying Gaussian mixture model and starts with an over segmented initial estimation which is adaptively modified until the optimal number of homogeneous texture segments is reached. The performance of the presented method is extensively tested on the Prague segmentation benchmark using the commonest segmentation criteria and compares favourably with several leading alternative image segmentation methods.
test's
TBES
version 1.0
WEB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Shankar Rao, Hossein Mobahi, Allen Yang, Shankar Sastry and Yi Ma:   Natural Image Segmentation with Adaptive Texture and Boundary Encoding

Algorithm for unsupervised segmentation of natural images that harnesses the principle of minimum description length (MDL). The method is based on observations that a homogeneously textured region of a natural image can be well modeled by a Gaussian distribution and the region boundary can be effectively coded by an adaptive chain code. The optimal segmentation of an image is the one that gives the shortest coding length for encoding all textures and boundaries in the image, and is obtained via an agglomerative clustering process applied to a hierarchy of decreasing window sizes. The optimal segmentation also provides an accurate estimate of the overall coding length and hence the true entropy of the image.
kukacji1's
ASATCSI
version 1.1
BIB
DOC
f1: 1
f2: 0
f3: 1
f4: 0
Manik Varma, Andrew Zisserman:   A Statistical Approach to Texture Classification from Single Images

This algorithm classifies textures based on known database of textures (also generated in this agorithm) using various filters. Basic statistic evaluation is used to determine which texture is examined. Texture should be identified regardless it's rotation and illumination.
novako20's
LBP-HF
SRC
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Ondrej Novak:   Local Binary Pattern Histogram Fourier Features

Paper: Rotation Invariant Image Description with Local Binary Pattern Histogram Fourier Features
Authors: Timo Ahonen, Jiri Matas, Chu He, Matti Pietikainen
Published: 2009
mevenkamp's
PCA-MS
BIB
WEB
SRC
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Niklas Mevenkamp, Benjamin Berkels:   Variational Multi-Phase Segmentation using High-Dimensional Local Features

A variational multi-phase segmentation framework based on the Mumford-Shah energy, combined with PCA-based dimension reduction is used to segment color or gray-value images into regions of different structure identified by high-dimensional features, such as local spectral histograms (for Texture) and localized Fourier transforms (for Crystals).
v-andrearczyk's
FCNT
version s/u
WEB
DOC
f1: 1
f2: 0
f3: 1
f4: 0
Vincent Andrearczyk, Paul F. Whelan:   Fully Convolutional Network for Texture

Versions: supervised / unsupervised
emkay's
A3M
version 1.0
BIB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Martin Kiechle, Martin Storath, Andreas Weinmann, Martin Kleinsteuber:   Model-based learning of local image features for unsupervised texture segmentation

Features that capture well the textural patterns of a certain class of images are crucial for the performance of texture segmentation methods. The manual selection of features or designing new ones can be a tedious task. Therefore, it is desirable to automatically adapt the features to a certain image or class of images. Typically, this requires a large set of training images with similar textures and ground truth segmentation. In this paper, we propose a framework to learn features for texture segmentation when no such training data is available. The cost function for our learning process is constructed to match a commonly used segmentation model, the piecewise constant Mumford-Shah model. This means that the features are learned such that they provide an approximately piecewise constant feature image with a small jump set. Based on this idea, we develop a two-stage algorithm which first learns suitable convolutional features and then performs segmentation. We note that the features can be learned from a small set of images, from a single image, or even from image patches. The proposed method achieves a competitive rank in the Prague texture segmentation benchmark, and it is effective for segmenting histological images.
huangyuan's
EWT-FCNT
BIB
DOC
f1: 1
f2: 0
f3: 0
f4: 0
Yuan Huang, Fugen Zhou, Jérôme Gilles:   Empirical curvelet based Fully Convolutional Network for supervised texture image segmentation

kuzelon3's
LBP
version 1.0
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Ondřej Kužela:   Local Binary Patterns

MI-ROZ course project
barusmar's
wold
version 1
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Martin Barus:   wold decomposition

School project
lorenpe2's
Three-Patch Code
version 1
f1: 0
f2: 0
f3: 0
f4: 0
Wolf, L., Hassner, T., Taigman, Y.:   Wolf, L., Hassner, T., Taigman, Y. (2008) Descriptor based methods in the wild. In ECCV workshop on faces in real-life images: Detection, alignment, and recognition.

Wolf, L., Hassner, T., Taigman, Y. (2008) Descriptor based methods in the wild. In ECCV workshop on faces in real-life images: Detection, alignment, and recognition.
frzn's
texNCUT
f1: 0
f2: 0
f3: 1
f4: 0
Farzaneh Alizadeh, Nader Karimi, Niloofar Gheissari:   Textural image segmentation using Normalized Cut

TexNCUT use from Texture features and a graph based image segmentation method(Ncut) for textural image segmentation. our algorithm employ super-pixels to increase speed and efficiency.

In TexNCUT the number of regions for the evaluated partitions was manually set to the number of regions in the ground truth partitions.
py's
CGCHi
version 0.0.1
f1: 0
f2: 0
f3: 0
f4: 0
Ruhallah Amandi, Mohammad Farhadi:   Combined Graph Cut based segmentation with histogram information on regions

Segmentation is one the more challenging problems in image processing. All segmentation algorithms must answer the three main questions, coherent definition-method to encode coherent in mathematical notion and give the final solution for such system. In this work we use the combination of global and local coherent. Find the sufficient number of clusters do by using histograms and probability theory, on the next part we use the metric space strategies to model local intensity feature of input image. Some problems in this method are from two main sources, wrong number of cluster estimation and the other one is the modeling method failures.

Multiregion Image Segmentation by Parametric Kernel Graph Cuts

Histogram clustering for unsupervised image segmentation
Mori's
Morfological Opening Filter
version 1
f1: 0
f2: 0
f3: 0
f4: 0
Veronika Maurerova:   Morfological Opening Filter

Morphological opening on an image is defined as an erosion followed by a dilation. Opening can remove small bright spots (i.e. “salt”) and connect small dark cracks.

(source: http://scikit-image.org/docs/dev/auto_examples/applications/plot_morphology.html)
hakmart1's
LevelSet
version 2
f1: 0
f2: 0
f3: 0
f4: 0
Martin Hak, Jiří Kovačič, Vojtěch Stránský:   LevelSet segmenter

Project for ROZ. Implementation of segmenter based on levelset method.
yuanj's
FSEG
version 1.0
WEB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Jiangye Yuan:   FSEG

The factorization based texture segmentation algorithm is applied.

No human interaction or prior information is needed.
palkoigo's
S/U-SRM
version 0.5
BIB
SRC
f1: 0
f2: 0
f3: 1
f4: 0
Igor Palkoci:   Statistical Region Merging - Supervised/Unsupervised version

david0432's
Cooperative Mum-Shah
version 1.1
f1: 0
f2: 0
f3: 1
f4: 0
David Perez, Felipe Calderero:   Cooperative Region Merging

Versions:  weighted memory / simple memory
tomas.duda's
MI-ROZ
version 0.1
f1: 0
f2: 0
f3: 0
f4: 0
Tomáš Duda:   Rozpoznávání - test

Test runs.
josuegalindo's
Quad
version 1.0
f1: 1
f2: 0
f3: 1
f4: 0
Jg:   Quad texture

Segments 4 textures
zitnyjak's
calderero
version KL un-/weighted
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Region Merging Techniques Using Information Theory Statistical Measures Segmentation (KL un-/weighted)

Segmentation algorithm based on paper "Region Merging Techniques Using Information Theory Statistical Measures" by Calderero and Marques - area un-/weighted with Kullback-Leibler merging criterion. Binary is 64bit.
ebasaeed's
learnFeaturesRS
version 1
WEB
DOC
f1: 1
f2: 0
f3: 1
f4: 0
Tuia, D.; Volpi, M. ; Dalla Mura, M. ; Rakotomamonjy, A. ; Flamary, R.:   Automatic Feature Learning for Spatio-Spectral Image Classification With Sparse SVM

magdafried's
CBP
version 0.0.1
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Centralized Binary Patterns

Centralized Binary Patterns Embedded with Image Euclidean Distance for Facial Expression Recognition
Published in: Natural Computation, 2008. ICNC '08, page(s): 115 - 119
felipecalderero's
GSRM MARKOV unsup.
version BHAT/KL area-weighted/-unweighted
WEB
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Felipe Calderero:   General statistical region merging MARKOV - unsupervised - 10 bins

felipecalderero's
GSRM MARKOV sup.
version BHAT/KL area-weighted/-unweighted
WEB
DOC
f1: 0
f2: 0
f3: 1
f4: 0
Felipe Calderero:   General statistical region merging MARKOV - supervised - 10 bins

ondrasim's
FIT_ROZ_15
version v_1.0.0
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Šimon Ondráček:   FIT_ROZ_15

Local difference
haurvojt's
WLD
version 1.10
f1: 0
f2: 0
f3: 0
f4: 0
Weber Local Descriptor with UHoG

The algorithm combines Weber excitation with unsigned Histogram of Gradients.

1.0:plain with detection window 9x9
1.1:plain with detection window 7x7
1.2:plain with detection window 5x5
1.3:histograms of WLD collected on every color plane instead of only the one with biggest histogram
1.4: added normalized histogram of color in the detection window
1.6: changed color to HSV space
1.7: changed back to RGB, added normalized location of pixel to feature vector
1.8: removed location, added pixel color value (with normalization)
1.10: smaller color detection window
labanjak's
Affinity Propagation
version 1.1.0
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Jakub Labant, Andrej Kudinov, Dausheyev Jorj-Mukhammed:   Clustering by passing messages between data points

babicpe1's
Surf
DOC
f1: 0
f2: 0
f3: 0
f4: 0
Surf

SURF alg based on http://www.vision.ee.ethz.ch/~surf/eccv06.pdf
hajkokla's
Haralick
f1: 0
f2: 0
f3: 0
f4: 0
Haralick textural features

[FIT2016] Monospectral pixelwise features
dlapavoj's
ROZ_LBP
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Vojtech Dlapal:   Local binary patterns by Vojtech Dlapal

hejlfran's
CTEX
version 1.0
f1: 0
f2: 0
f3: 0
f4: 0
Dana E. Ilea and Paul F. Whelan:   An Adaptive Unsupervised Segmentation Algorithm Based on Color-Texture Coherence

stanej14's
Textons
f1: 1
f2: 0
f3: 0
f4: 0
Manik Varma and Andrew Zisserman:   A Statistical Approach to Texture Classification from Single Images

kubeljit's
Run len matrix
version 2.0
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Run len matrix

Přidány všechny features popsané v článku - Galloway, Chu a Dasarathy&Holde.
malyja's
scale_aap_segm
version 1.0
f1: 0
f2: 0
f3: 0
f4: 0
Jan Malý:   The Scale of a Texture and its Application to Segmentation

tauchkri's
LDP
version 1.0
f1: 0
f2: 0
f3: 0
f4: 0
tauchkri:   Local Derivative Pattern

malicto1's
Gabor features
f1: 0
f2: 0
f3: 0
f4: 0
Gabor features

Test - vyzkouseni nahrani vysledku s puvodni metodou gaussian blur bez jakekoliv upravy.
daicav's
SegTexCol
version 1
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Segmentation using texture and colour information

connetwork12's
conn
f1: 1
f2: 0
f3: 0
f4: 0
conn

Versions: gl/col - noblur/blur3/blur6
martinkersner's
Genetic Algorithm
version 1.0.
f1: 0
f2: 0
f3: 0
f4: 0
Genetic Algorithm

Multi-thresholding
siddharth's
My algo1
version 0.1
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Siddharth Kumar:   My algo1

horkyja6's
FITsegmenter
version 1.3 v11x
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Jan Horký:   FITsegmenter

novotm@fit.cvut's
Gabor features
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Gabor features

alexandliutao@gmail.com's
==
f1: 0
f2: 0
f3: 0
f4: 0
FSEG_nocolor / FSEG_onlycolor
sparse+unsupervised / filter+sparse / filter+sparse+process
atom30t23 / atom20t14 / atom20t6
my_unsupervised / multiscale

xiaofang's
celine.chinoise
f1: 0
f2: 0
f3: 1
f4: 0
sas_gmm(color+tensor), sas_gmm(color) [c=7]
weighted_color_patch, gmm_sift [c=7]
sas, gmm_lrr, sas_gmm_withoutSparseCoding
SR_multifeat [all]

wallejak's
Gravitational model
version 1.3
f1: 0
f2: 0
f3: 0
f4: 0
A simplified gravitational model to analyze texture roughness

icpr2014_test's
icpr2014_alg
f1: 0
f2: 0
f3: 0
f4: 0
** auxiliary - for uploading icpr2014 contest results **

test's
test
f1: 0
f2: 0
f3: 1
f4: 0
** auxiliary - for uploading miscellaneous results **

juzlomar's
ELTFS
f1: 0
f2: 0
f3: 0
f4: 0
Enhanced Local Texture Feature Sets

polakluk's
MI-ROZ sem
f1: 0
f2: 0
f3: 1
f4: 0
Zitny, Polak, Makara:   MI-ROZ sem

th's
==
f1: 0
f2: 0
f3: 0
f4: 0
texture2015 / multi2015 / semi

MikeDoni's
ICG_Segmenter
version 1.0
f1: 0
f2: 0
f3: 0
f4: 0
Michael Donoser:   ICG_Segmenter

rumanjak's
CBP
f1: 0
f2: 0
f3: 0
f4: 0
Centralised Binary Pattern

palicand's
voting max
version 0.1
f1: 0
f2: 0
f3: 1
f4: 0
Voting Maximization

alexama1's
ROZ test
version 0.0.1
f1: 1
f2: 1
f3: 1
f4: 1
Marek Alexa:   ROZ test

liutoole's
Laws Filter Masks
version 1.1
SRC
f1: 0
f2: 0
f3: 0
f4: 0
Laws Filter Masks

koppmaty's
CONVOLUTION
version 2.17
SRC
f1: 0
f2: 0
f3: 0
f4: 0
CONVOLUTION

svehlja5's
Histograms of Oriented Gradients
f1: 0
f2: 0
f3: 0
f4: 0
Histograms of Oriented Gradients

prochm35's
Dominant Neighborhood Structure
version 12
f1: 0
f2: 0
f3: 0
f4: 0
Dominant Neighborhood Structure

rohit.iiith's
LBP based texture segmentation
f1: 0
f2: 0
f3: 0
f4: 0
LBP based texture segmentation

kratolu3's
Dominant Local Binary Patterns
f1: 0
f2: 0
f3: 0
f4: 0
Dominant Local Binary Patterns

ebasaeed's
WS
version Paper (mscnn; boosting; mean, no Merging) v27
f1: 1
f2: 1
f3: 0
f4: 0
WS

lingxiu's
ICM
version matlab
SRC
f1: 1
f2: 0
f3: 0
f4: 0
ICM

rohit.iiith's
Graph based Segmentation
f1: 0
f2: 0
f3: 0
f4: 0
Graph based Segmentation

veselj38's
Histogram ratio features
version 0.1
f1: 0
f2: 0
f3: 0
f4: 0
Histogram ratio features

polkennel's
DTCWT_texton_SVM
f1: 1
f2: 0
f3: 0
f4: 0
DTCWT_texton_SVM

feantury's
KMeans
version 1.0
SRC
f1: 1
f2: 0
f3: 1
f4: 0
KMeans

ruhiravichandran's
==
version 1.1
SRC
f1: 0
f2: 0
f3: 0
f4: 0
guloleg's
HOG
version 1
SRC
f1: 1
f2: 0
f3: 1
f4: 0
HOG

tothmatu's
HASC
SRC
f1: 0
f2: 0
f3: 0
f4: 0
HASC

nemecst4's
segm_tut_101019
version 1
f1: 0
f2: 0
f3: 0
f4: 0
segm_tut_101019

lana's
tut2
version 5
SRC
f1: 0
f2: 0
f3: 1
f4: 0
tut2

bilikdav's
ROZ-bilikdav
version 1
f1: 1
f2: 0
f3: 1
f4: 0
ROZ-bilikdav

vanclmil's
helloworld
f1: 0
f2: 0
f3: 0
f4: 0
helloworld

strejivo's
segmenter1
version 1.0
f1: 0
f2: 0
f3: 0
f4: 0
segmenter1

xaos's
LBP + EM
f1: 0
f2: 0
f3: 0
f4: 0
LBP + EM

huangyuan's
unet
version psp21ewt
f1: 1
f2: 0
f3: 0
f4: 0
unet

1354884112@qq.com's
ltsparse
f1: 0
f2: 0
f3: 0
f4: 0
ltsparse

xaos's
AR2D+EM
f1: 0
f2: 0
f3: 0
f4: 0
AR2D+EM

mahelpet's
noname
f1: 1
f2: 0
f3: 0
f4: 0
noname

dteney's
test
f1: 0
f2: 0
f3: 1
f4: 0
test

yuanj's
RS
f1: 0
f2: 0
f3: 0
f4: 1
RS

slavojir's
Wold
f1: 0
f2: 0
f3: 0
f4: 0
Wold

hejnapet's
SIFT
version 14
f1: 0
f2: 0
f3: 0
f4: 0
SIFT

richtrad's
3DFT
version 0.1
f1: 0
f2: 0
f3: 0
f4: 0
3DFT

pirojan's
MI-ROZ
f1: 0
f2: 0
f3: 0
f4: 0
MI-ROZ

macaond3's
Roz
f1: 0
f2: 0
f3: 0
f4: 0
Roz

malekva1's
alg1
f1: 0
f2: 0
f3: 0
f4: 0
alg1

cernyon3's
ALG
f1: 0
f2: 0
f3: 0
f4: 0
ALG

brokejan's
Colors
version 1
f1: 1
f2: 0
f3: 0
f4: 0
Colors

dressmar's
tmp
version 1
f1: 0
f2: 0
f3: 0
f4: 0
tmp