A combined neural network approach for texture classification

doi:10.1016/0893-6080(95)00013-P

Neural Networks

Volume 8, Issue 6, 1995, Pages 975-987

https://doi.org/10.1016/0893-6080(95)00013-P Get rights and content

Abstract

In this article, we present a two-stage neural network structure that combines the characteristics of self-organizing map (SOM) and multilayer perceptron (MLP) for the problem of texture classification. The texture features are extracted using a multichannel approach. The channels comprise of a set of Gabor filters having different sizes, orientations, and frequencies to constitute N-dimensional feature vectors. SOM acts as a clustering mechanism to map these N-dimensional feature vectors onto its M-dimensional output space, where in our experiments, the value of M was taken as two. This, in turn, forms the feature space from which the features are fed into an MLP for training and subsequent classification. It is shown that the disadvantage of using Gabor filters in texture analysis, namely, the higher dimensionality of the Gaborian feature space, is overcome by the reduction in the dimensionality of the feature space achieved by SOM. This results in a significant reduction in the learning time of MLP and hence the overall classification time. It is found that this mechanism increases the interclass distance (average distance among the vectors of different classes) and at the same time decreases the intraclass distance (average distance among the vectors of the same class) in the feature space, thereby reducing the complexity of classification. Experiments were performed on images containing tiles of natural textures as well as image data from remote sensing.

References (39)

J.M. Coggins et al.
A spatial filtering approach to texture analysis
Pattern Recognition Letters
(1985)
J.G. Daugman
Two-dimensional spectral analysis of cortical receptive field profiles
Vision Research
(1980)
R.L. De Valois et al.
Spatial frequency selectivity of cells in macaque visual cortex
Vision Research
(1982)
L.D. Jacobson et al.
Joint spatial/spatialfrequency representations
Signal Processing
(1988)
A.K. Jain et al.
Unsupervised texture segmentation using Gabor filters
Pattern Recognition
(1991)
J.M. Keller et al.
Texture description and segmentation through fractal geometry
Computer Vision, Graphics and Image Processing
(1989)
Ari Visa
A texture classifier based on neural network principles
R. Bajcsy
Computer description of textured surfaces
D.H. Berger
Texture as a discriminant of crops on radar imagery
IEEE Transactions on Geoscience and Electronics
(1970)
A.C. Bovik et al.
Computational texture analysis using localized spatial filtering

A.C. Bovik et al.

Multichannel texture analysis using localized spatial filters

IEEE Transactions on Pattern Analysis and Machine Intelligence

(1990)

F.W. Campbell et al.

Application of Fourier analysis of the visibility of gratings

Journal of Physiology (London)

(1968)

R. Chellappa et al.

Texture segmentation with neural networks

G.B. Coleman et al.

Image segmentation by clustering

J.G. Daugman

Uncertainty relation for resolution in space, spatial-frequency, and orientation optimized by twodimensional visual cortical filters

Journal of the Optical Society of America A

(1985)

J.G. Daugman

Complete discrete 2-D Gabor transforms by neural network for image analysis and compression

IEEE Transactions on Acoustics, Speech and Signal Processing

(1988)

F.A. DeCosta et al.

Neural network recognition of textured images using third order cumulants as functional links

N.R. Dupaguntla et al.

A neural network architecture for texture segmentation and labeling

D. Gabor

Theory of communication

Journal of the IEE

(1946)

Cited by (32)

A global-best harmony search based gradient descent learning FLANN (GbHS-GDL-FLANN) for data classification
2016, Egyptian Informatics Journal
Citation Excerpt :
The remaining part of this paper is organized as follows: Preliminaries in Section 4, proposed method in Section 5, experimental setup in Section 6, simulation results and performance comparisons in Section 7, proof of statistical significance in Section 8, conclusion in Section 9 and references. The Functional Link Artificial Neural Network (FLANN) [240] is a class of Higher Order Neural Networks that make use of higher combination of its inputs [241,242] and has been successfully used in many applications such as pattern recognition [243,244], classification [245–247], channel equalization [248], system identification [249–253] and prediction [254]. Even if it has a single-layer network, still it is capable to handle nonlinear separable classification task as compared to MLP.
While dealing with real world data for classification using ANNs, it is often difficult to determine the optimal ANN classification model with fast convergence. Also, it is laborious to adjust the set of weights of ANNs by using appropriate learning algorithm to obtain better classification accuracy. In this paper, a variant of Harmony Search (HS), called Global-best Harmony Search along with Gradient Descent Learning is used with Functional Link Artificial Neural Network (FLANN) for classification task in data mining. The Global-best Harmony Search (GbHS) uses the concepts of Particle Swarm Optimization from Swarm Intelligence to improve the qualities of harmonies. The problem solving strategies of Global-best Harmony Search along with searching capabilities of Gradient Descent Search are used to obtain optimal set of weight for FLANN. The proposed method (GbHS-GDL-FLANN) is implemented in MATLAB and compared with other alternatives (FLANN, GA based FLANN, PSO based FLANN, HS based FLANN, Improved HS based FLANN, Self Adaptive HS based FLANN, MLP, SVM and FSN). The GbHS-GDL-FLANN is tested on benchmark datasets from UCI Machine Learning repository by using 5-fold cross validation technique. The proposed method is analyzed under null-hypothesis by using Friedman Test, Holm and Hochberg Procedure and Post-Hoc ANOVA Statistical Analysis (Tukey Test & Dunnett Test) for statistical analysis and validity of results. Simulation results reveal that the performance of the proposed GbHS-GDL-FLANN is better and statistically significant from other alternatives.
A self adaptive harmony search based functional link higher order ANN for non-linear data classification
2016, Neurocomputing
Citation Excerpt :
Notwithstanding the fact that it has a single-layer network, it is capable to handle non-linear separable classification task as compared to MLP. Functional link artificial neural network (FLANN) [38] utilizes the higher combination of its inputs [39,40] and has been successfully applied in a wide spectrum of applications such as pattern recognition [41,42], classification [43–45], channel equalization [46] , system identification [47–52], etc. Almost all the higher order ANNs (HONNs) including functional link higher order ANN (FLANN) are sensitive to random initialization of weight and rely on the learning algorithm adopted.
In the data classification process involving higher order ANNs, it’s a herculean task to determine the optimal ANN classification model due to non-linear nature of real world datasets. To add to the woe, it is tedious to adjust the set of weights of ANNs by using appropriate learning algorithm to obtain better classification accuracy. In this paper, an improved variant of harmony search (HS), called self-adaptive harmony search (SAHS) along with gradient descent learning is used with functional link artificial neural network (FLANN) for the task of classification in data mining. Using its past experiences, SAHS adjusts the harmonies according to the maximum and minimum values in the current harmony memory. The powerful combination of this unique strategy of SAHS and searching capabilities of gradient descent search is used to obtain optimal set of weights for FLANN. The proposed method (SAHS–FLANN) is implemented in MATLAB and the results are contrasted with other alternatives (FLANN, GA based FLANN, PSO based FLANN, HS based FLANN, improved HS based FLANN and TLBO based FLANN). To illustrate its effectiveness, SAHS–FLANN is tested on various benchmark datasets from UCI machine learning repository by using 5-fold cross validation technique. Under the null-hypothesis, the proposed method is analyzed by using various statistical tests for statistical correctness of results. The performance of the SAHS–FLANN is found to be better and statistically significant in comparison with other alternatives. The SAHS–FLANN differs from HS–FLANN (HS based FLANN) by the elimination of constant parameters (bandwidth and pitch adjustment rate). Furthermore, it leads to the simplification of steps for the improvisation of weight-sets in IHS–FLANN (improved HS based FLANN) by incorporating adjustments of new weight-sets according to the weight-sets with maximum and minimum fitness.
Neural network-based text location in color images
2001, Pattern Recognition Letters
This paper proposes neural network-based text locations in complex color images. Texture information extracted on several color bands using neural networks is combined and corresponding text location algorithms are then developed. Text extraction filters can be automatically constructed using neural networks. Comparisons with other text location methods are presented; indicating that the proposed system has a better accuracy.
Geostatistical classification for remote sensing: An introduction
2000, Computers and Geosciences
Traditional spectral classification of remotely sensed images applied on a pixel-by-pixel basis ignores the potentially useful spatial information between the values of proximate pixels. For some 30 years the spatial information inherent in remotely sensed images has been employed, albeit by a limited number of researchers, to enhance spectral classification. This has been achieved primarily by filtering the original imagery to (i) derive texture ‘wavebands’ for subsequent use in classification or (ii) smooth the imagery prior to (or after) classification. Recently, the variogram has been used to represent formally the spatial dependence in remotely sensed images and used in texture classification in place of simple variance filters. However, the variogram has also been employed in soil survey as a smoothing function for unsupervised classification. In this review paper, various methods of incorporating spatial information into the classification of remotely sensed images are considered. The focus of the paper is on the variogram in classification both as a measure of texture and as a guide to choice of smoothing function. In the latter case, the paper focuses on the technique developed for soil survey and considers the modification that would be necessary for the remote sensing case.
Automatic estimation of crowd density using texture
1998, Safety Science
This paper considers the role of automatic estimation of crowd density and its importance for the automatic monitoring of areas where crowds are expected to be present. A new technique is proposed which is able to estimate densities ranging from very low to very high concentration of people, which is a difficult problem because in a crowd only parts of people's body appear. The new technique is based on the differences of texture patterns of the images of crowds. Images of low density crowds tend to present coarse textures, while images of dense crowds tend to present fine textures. The image pixels are classified in different texture classes and statistics of such classes are used to estimate the number of people. The texture classification and the estimation of people density are carried out by means of self organising neural networks. Results obtained respectively to the estimation of the number of people in a specific area of Liverpool Street Railway Station in London (UK) are presented.
A hybrid neural network model in handwritten word recognition
1998, Neural Networks
A hybrid neural network model is developed and applied to handwritten word recognition. The word recognition system requires a module that assigns character class confidence values to segments of images of handwritten words. The module must accurately represent ambiguities between character classes and assign low confidence values to a wide variety of non-character segments resulting from erroneous segmentations. The proposed hybrid neural model is a cascaded system. The first stage is a self-organizing feature map algorithm (SOFM). The second stage maps distances into allograph membership values using a gradient descent learning algorithm. The third stage is a multi-layer feedforward network (MLFN). The new system performs better than the baseline system. Experiments were performed on a standard test set from the SUNY/USPS Database.

View all citing articles on Scopus

View full text

Contributed articleA combined neural network approach for texture classification

Abstract

Pattern Recognition Letters

Vision Research

Vision Research

Signal Processing

Pattern Recognition

Computer Vision, Graphics and Image Processing

A texture classifier based on neural network principles

Computer description of textured surfaces

Texture as a discriminant of crops on radar imagery

IEEE Transactions on Geoscience and Electronics

Computational texture analysis using localized spatial filtering