Rama Chellappa and S. Chatterjee. Classification of textures using Gaussian Markov random fields. IEEE Trans. ASSP, 33:959--963, 1985.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....as a textural pattern different from the normal pattern [1] A number of approaches to solve the texture classification problem have been developed over the years. Early research work was based on the first order and second order statistics of texture [2] 5] Gaussian Markov random fields [6], 7] fractal models [8] 12] and local linear transforms [13] 16] Recently, some modern methods have been developed, such as multichannel methods, multiresolution analysis, Gabor filters, and the wavelet transform [17] 27] Many of these approaches have provided good results in different ....

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. Acoust., Speech, Signal Processing, vol. ASSP-33, pp. 959--963, Aug. 1985.

....This approach requires the number and type of textures to be known in advance. That is, a set of training textures are used to formalise the criteria by which the texture models become unique from each other, but not necessarily unique from any other textures not included in the training set [27, 40, 38, 106]. These conventional models need only capture enough textural characteristics to classify the textures in the training set, via discriminant analysis [60] This approach is adequate if the image undergoing texture segmentation and classification is known to contain only those textures from the ....

....correct estimation of the MRF very di#cult. To make the estimation process a little easier it is common to constrain an MRF model as an auto model, which only contains pairwise interactions. There are still many variations of the auto model [17] e.g. Derin Elliot model [54] auto normal model [40, 48], and the auto binomial model [17, 50] These models were able to capture micro textures well, but failed with regular and inhomogeneous textures. MRF models have been applied to various image processing applications such as texture synthesis [50] texture classification [40, 120] image ....

[Article contains additional citation context not shown here]

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP--33, no. 4, pp. 959--963, 1985.

....which we have no prior knowledge of the constituent texture types. This technique can therefore be used to find a specific texture in a background of unknown textures. I. Introduction The process of classifying textures in an image usually requires prior knowledge of all textures that may occur [1]. Where this is known, texture models need capture only su#cient characteristics to discriminate between the set of known textures, and then discriminant analysis for instance may be used [2] However, for images where not all textural types are known, a texture model needs to capture all relevant ....

Rama Chellappa and Shankar Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP--33, no. 4, pp. 959--963, 1985.

....current techniques require the number and type of textures to be a prior known. That is, they use a set of training textures to formalise the criteria by which the texture models become unique from each other, but not necessarily unique from other textures not included in the training set [12] [10], 7] 29] These conventional models need only capture enough textural characteristics to classify the set of textures via discriminant analysis [17] This approach is adequate if the image undergoing texture segmentation and classification is known to contain only textures which were modelled. ....

....while retaining the unique characteristics of the texture [45] VI. Multiscale Unsupervised Texture Recognition MRF models have mainly been used for the supervised classification of texture, for which a library of pre modelled textures must exist in order for discriminant analysis to be used. [10], 15] 23] 12] 9] However, this approach would be cumbersome for SAR images of the Earth s terrain as they contain a myriad of di#erent texture types, too many to be able build a library of pre modelled textures such that discriminant analysis maybe performed on an arbitrary image. We ....

Rama Chellappa and Shankar Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP--33, no. 4, pp. 959--963, 1985.

....current techniques require the number and type of textures to be a prior known. That is, they use a set of training textures to formalise the criteria by which the texture models become unique from each other, but not necessarily unique from other textures not included in the training set. [7,11,12,31]. These conventional models need only capture enough textural characteristics to classify the set of textures via discriminant analysis [17] This approach is adequate if the image undergoing texture segmentation and classification is known to contain only textures which were modelled. If a ....

....while retaining the unique characteristics of the texture [47] VI. MULTISCALE UNSUPERVISED TEXTURE RECOGNITION MRF models have mainly been used for the supervised classification of texture, for which a library of pre modelled textures must exist in order for discriminant analysis to be used. [9,11,12,16,22]. However, this approach would be cumbersome for SAR images of the Earth s terrain as they contain a myriad of different texture types, too many to be able build a library of pre modelled textures such that discriminant analysis maybe performed on an arbitrary image. We present a new approach to ....

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov Ran- dom Fields," IEEE Transactions on Acoustics, Speech, and SignM Processing, ASSP-33, no. 4, pp. 959-963, 1985.

....developed in this paper, while many others have close ties to it. 2.4 Texture Discrimination Another problem of importance in computer vision and in other image processing applications is that of texture discrimination. One well known class of statistical texture models is that based on MRF s [50, 178, 71, 233]. For example, Figure 4 shows two synthetic MRF textures, one modeling pigskin and one sand. The problem of discriminating textures such as these given noisy measurements is a standard hypothesis testing problem whose solution hinges on the computation of the likelihood ratio for the two textures ....

....in Section 2.2 and developed more thoroughly in [225] As mentioned in Section 2, many textures, such as those shown in Figure 4, can be modeled e#ectively using MRF models. However, likelihood function computations for MRF s are highly nontrivial, so that suboptimal methods (such as those in [50]) are often used. Another approach starts with the simple and obvious statement that such an MRF model does not represent truth but rather is itself an idealization of real textures. As a result, it is reasonable to seek alternate models that lead to much simpler likelihood computations ....

[Article contains additional citation context not shown here]

R. Chellappa and S. Chatterjee. Classification of textures using Gaussian Markov random fields. IEEE Trans. on Acoustic Speech and Signal Processing, ASSP-33(4):959--963, August 1985. 67

....Texture synthesis, Unsupervised texture segmentation, Unsupervlsed texture recognition. I. INTRODUCTION MR, F models have mainly been used for the supervised classification of texture, for which a library of pre modelled textures must exist in order for discriminant analysis to be used. [3,4, 5,6,8]. However, this approach would be cumbersome for SAR, images of the Earth s terrain as they contain a myriad of different texture types, too many to be able build a library of pre modelled textures such that discriminant analysis maybe performed on an arbitrary image. We present a new approach to ....

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov Random Fields," IEEE Transactions on Acoustics, Speech, and Signal Processing, ASSP-33, no. 4, pp. 959-963, 1985.

....with a particular realisation # is defined as, D# ( 4 y z 7 B qt7 pHq V# ( 6) where y is the normalising constant or partition function, z 3 s: D# ( 7) Generally y is intractable both analytically and numerically. One exception is the Gaussian MRF model [5, 6] 4.1 Cliques Given a neighbourhood system , a clique is a set if every pair of distinct sites in neighbours. That is, given I 0J implies TF . The single site subset is also a clique. denote the set of cliques defined on with respect to , and let denote the local clique ....

R. Chellappa and S. Chatterjee. Classification of textures using Gaussian Markov random fields. IEEE Transactions on Acoustics, Speech, and Signal Processing, ASSP--33(4):959--963, 1985.

....of features extracted from the image, instead of the image itself. The Markov Random Field (MRF) models are one such example; see, e.g. References [6, 7] However, these models tend to be computationally expensive. Recently, De Bonet and Viola s proposed a flexible histogram approach[8, 9], where features are extracted at multiple image scales, with the resulting feature vectors treated as a set of independent samples drawn from a distribution. The distribution of feature vectors is then modeled using Parzen windows. Though they report good results, their model treats the feature ....

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. ASSP, vol. 33, pp. 959--963, 1985.

....sets. Index Terms Gabor filters, texture classification, wavelets. I. INTRODUCTION T HE SPECTRUM of texture analysis techniques ranges from those focusing on structural features to those emphasizing statistical modeling. In most statistically oriented techniques within the last 15 years [6], 11] 15] 16] 20] the image is modeled as a Markov random field (MRF) of pixels. In these approaches, the relationships between the intensities of neighboring pixels are statistically characterized. These methods have proven very effective for texture segmentation and classification. More ....

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random field models," IEEE Trans. Acoust., Speech, Signal Processing, vol. 33, pp. 959--963, Aug. 1985.

....There are several approaches that do not model the probability distribution on an image We thank Jeremy De Bonet and John Fisher for kindly answering questions about their work and experiments. This work supported by the United States Government. space, but motivated our work, e.g. MRF models [2, 3], the flexible histogram approach [4, 5] and multiscale stochastic processes [6] All of these methods seem to be well suited for modeling texture, but it is unclear how we might use them to capture the appearance of more structured objects. As in many other approaches, we model the distribution ....

Rama Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. ASSP, vol. 33, pp. 959--963, 1985.

....Also, affiliated with the Institute for Systems Research. Supported in part by Air Force grant No. F49620 92 J0130. 1 Introduction Random Fields have been successfully used to sample and synthesize textured images ( 14] 10] 24] 21] 17] 32] 12] 11] 15] 18] 9] 37] 6] [7], 8] Texture analysis has applications in image segmentation and classification, biomedical image analysis, and automatic detection of surface defects. Of particular interest are the models that specify the statistical dependence of the gray level at a pixel on those of its neighborhood. There ....

....the sampling process for generating synthetic textured images, and algorithms that yield an estimate of the parameters of the assumed random process given a textured image. Impressive results related to real world imagery have appeared in the literature ( 14] 17] 12] 18] 37] 6] [7], 8] However, all these algorithms are quite computationally demanding because they typically require on the order of G n arithmetic operations per iteration for an image of size n Theta n with G gray levels. The implementations known to the authors are slow and operate on images of size ....

R. Chellappa and S. Chatterjee. Classification of Textures Using Gaussian Markov Random Fields. IEEE Transactions on Acoustics, Speech, and Signal Processing, 33:959-- 963, August 1985.

No context found.

Rama Chellappa and S. Chatterjee. Classification of textures using Gaussian Markov random fields. IEEE Trans. ASSP, 33:959--963, 1985.

No context found.

R. Challappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. Acous. Speech. Signal Proc. 33, pp. 959--963, 1985.

No context found.

Rama Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. ASSP, vol. 33, pp. 959--963, 1985.

No context found.

R. Challappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. Acous. Speech. Signal Proc., vol. 33, pp. 959--963, 1985.

No context found.

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-33, no. 8, pp. 959--963, Aug. 1985.

No context found.

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP--33, no. 4, pp. 959--963, 1985.

No context found.

Chellapa R.,Chatterjee S., "Classification of Texture Using Gaussian Markov Random Fields" IEEE Trans on ASSP, vol. Assp-33,no. 4, pp. 959-963 Cohen F. S., Fan Z., Attali S. "Automated Inspection of Textile Fabrics Using Textural Models" IEEE Trans on PAMI, vol. Pami-13, no. 8, pp. 803-808

No context found.

Chellapa R.,Chatterjee S., "Classification of Texture Using Gaussian Markov Random Fields" IEEE Trans. on ASSP, vol. Assp-33, no. 4, pp. 959-963, August 1985

No context found.

Chellappa R, Chatterjee S. Classification of textures using gaussian markov random fields. IEEE Tr. Acoustic, Speech and Signal Processing 1985; 33: 959-963. 10

No context found.

R. Chellappa and S. Chatterjee. Classification of Textures using Gaussian Markov Random Fields. IEEE Transactions on Acoustics Speech and Signal Processing, 33:959--963, 1985.

No context found.

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Trans. Acoust., Speech, Signal Processing, vol. 33, pp. 959--963, Aug. 1985.

No context found.

R. Chellappa and S. Chatterjee, "Classification of textures using Gaussian Markov random fields," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP--33, no. 4, pp. 959--963, 1985.

No context found.

Chelappa, R., Chatterjee, S.: Classification of textures using Gaussian Markov random fields. IEEE Trans. Acoust., Speech, Signal Process. 33 (1985), pp. 959-963.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC