J. W. Modestino and J. Zhang, "A Markov random field model-based approach to image interpretation," IEEE Trans. Pattern Anal. Machine Intell., vol. 14, pp. 606-- 615, June 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... introduced in many fundamental issues of image analysis and computer vision such as image restoration [5] 14] edge detection [13] image segmentation [9] 13] computed tomography [11] surface reconstruction [9] 30] stereovision [2] motion analysis [18] 24] 37] or scene interpretation [33]. The mathematical framework is a statistical one: entities of interest in a given task are described by statistical models (Markov random fields) and Bayesian estimation theory is used to extract the relevant information from the observed images. By defining comprehensive global statistical ....

J.W. Modestino and J. Zhang, "A Markov random field modelbased approach to image interpretation," IEEE Trans. Pattern Anal. Machine Intell., vol. 14, no. 6, pp. 606-615, 1992.

....sought to make use of knowledge based (expert) systems for the sake of inference. The early work in this respect includes that by Ohta [1] Prasannappa, et al. [2] and McKeown, et al. 3] A probabilistic approach to domain independent image inter pretation was attempted by Modestino and Zhang [4]. Their approach was to model the segments of a segmented image as the nodes of an adjacency graph and their labels (interpretations) given the feature measurements (i.e. the a posteriori distribution) as a Markov Random Field (MRF) over the graph. Domain knowl edge was incorporated by ....

....Geman [5] was used to find the MAP estimate of the MRF which corresponds to the best interpretation given the features and spatial constraints. Recently Kim and Yang [6] demonstrated the use of error backpropagation networks to learn the various clique functions in the MRF framework proposed in [4]. The generality of an MRF model allows us to define a particu lar p.d.f. over the adjacency graph and obtain its MAP estimate. There is another p.d.f. which in its generality is amenable to similar treatment, namely, the p.d.f. of Bayesian networks. They have been used before for building expert ....

[Article contains additional citation context not shown here]

J.W. Modestino and J. Zhang, "A Markov random field modelbased approach to image interpretation," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 14, no. 6, pp. 606-615, June 1992.

.... of MRF applications, for instance, can be found in the monographs [20, 10] It is important to note that approaches which use MRF based modeling solve mostly low level image processing problems [21, 20] Only a few authors consider the high level vision task of object recognition using MRF [19, 11, 13]. And there is still the open question of defining an appropriate neighborhood system and energy function automatically; also the issue of self occlusion is not solved yet. The first approach to generate probabilistic models based on MRF automatically from observations was published in [21] The ....

J.W. Modestino and J. Zhang. A Markov random field model--based approach to image interpretation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14(6):606--615, June 1992.

.... applications can be found in literature (see for instance the monographs [26, 12] It is important to note that approaches which use MRF based models solve mostly low level image processing problems [27, 26] Only a few authors consider the high level vision task of object recognition using MRF [25, 13, 16]. This is due to the fact that high level vision tasks have to be generally modeled by irregular neighborhood systems, 2 which are mostly defined by means of heuristic distances, generally feature dependent. Consider for instance 3D object recognition as application problem: if the chosen ....

J.W. Modestino and J. Zhang. A Markov random field model--based approach to image interpretation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14(6):606-- 615, June 1992.

....graphical models can be subdivided into two broad categories depending on whether the edges of the associated graph are directed or undirected. Undirected edges are typical in problems where interactions are considered to be symmetric, such as in statistical mechanics or image processing [24, 17, 31]. In the undirected case, in one form or another, these models are called random Markov fields, undirected probabilistic independence networks, Boltzmann machines, Markov networks, and log linear models. Directed models are used in cases where interactions are not symmetric and reflect causal ....

.... corresponds to the notion of Markov random field, or Markov network, or probabilistic independence network, or, in a slightly different context, Boltzmann machine [24, 1] Symmetric interaction models are typically used in statistical mechanics for instance, Ising models and image processing [17, 31], where associations are considered to be more correlational than causal. 8.2.1 Markov Properties A Markov random field on a graph G is characterized by any one of the following three equivalent Markov independence properties. The equivalence of these properties is remarkable, and its proof is ....

J. M. Modestino and J. Zhang. A Markov random field model-based approach to image interpretation. IEEE Trans. Pattern Anal. Machine Intell. , 14:606--615, 1992.

....model based information is needed such as the intrinsic grouping properties of the geometric shapes used in direct matching methods and the relational and structural cues of the model. In fact, relational approaches have been successfully used in computer vision systems for interpretation [26], 32] and shape description [17] of natural images. Direct matching between the 3 D or 2 D models and the extracted image features has been used for recognizing rigid objects with strong viewpoint invariant features when the initial estimate of the 5 viewing position is close to the real one ....

....solution. In addition, the solution is compromised by the SDE boundary conditions. Finally, structural information seems important for recognition and interpretation in human vision [40] 23] This may explain the success of the approaches based upon relational models for image interpretation [26], 3] 12] In addition, relational and structural knowledge of the considered scene can be helpful for low level feature organization. Consequently, we believe that such high level information can provide structural constraints in a multiple curve matching approach. In this paper, we wish to ....

J.W. Modestino and J. Zhang, "A Markov Random Field Model-Based Approach to Image Interpretation," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 14, no. 6, pp. 606-615, 1992.

....estimated. For instance, global energy functions have been successfully introduced in image restoration [4, 9, 16] edge detection [14] image segmentation [14] stereovision [2] computed tomography, surface reconstruction [11] visual motion analysis [7, 8, 20, 26, 31] and scene interpretation [30]. However, minimizing a global energy function is often an intricate problem: the number of possible label configurations is generally very large and the global energy function may exhibit many local minima. Computationally demanding stochastic relaxation algorithms are generally necessary to ....

J.W. MODESTINO and J. ZHANG. -- A Markov Random Field modelbased approach to image interpretation. -- IEEE Trans. Pattern Anal. Machine Intell., Vol 14, No 6: pages 606--615, June 1992.

....expressed by local prior models. However, the use of Bayesian theory makes it possible to apply more general models. Indeed global prior models can be used. In these cases, Bayesian inference has recently proved to be also successful for high level image analysis tasks such as scene interpretation [9], shape matching with deformable template based methods [10] or classification problems [11] The Bayesian methodology allows to combine a prior model that expresses the available prior information on the hidden variables, or the solution to be estimated, along with a statistical model describing ....

J.W Modestino and J. Zhang. A Markov random field Model-based approach to image interpretation. IEEE Trans. on Pattern Analysis and Machine Intelligence, PAMI-14(6):601--615, 1992.

....be estimated. For instance, global energy functions have been successfully introduced in image restoration [4, 10, 18] edge detection [16] image segmentation [16] stereovision [2] computed tomography, surface reconstruction [12] visual motion analysis [8, 9, 21, 29] and scene interpretation [35]. However, minimizing a global energy function is often an intricate problem: the number of possible label configurations is generally very large and the global energy function may exhibit many local minima. Computationally demanding stochastic relaxation algorithms are generally necessary to ....

J.W. MODESTINO and J. ZHANG. -- A Markov Random Field model-based approach to image interpretation. -- IEEE Trans. Pattern Anal. Machine Intell., Vol. 14, No 6: pages 606--615, June 1992.

.... have been used in a wide range of applications including image restoration [3, 21] edge detection [20] luminance and texture segmentation [15, 17, 20, 23, 36] stereovision [1] computed tomography, surface reconstruction [11] visual motion analysis [5, 6, 27, 34] and scene interpretation [37]. Energy functions involve generally two components, one of which expresses the interaction between the hidden labels and the observations and the other which encodes constraints on the desired solution [20] The choice of these energy functions is either heuristic or may be guided by a ....

J.W. MODESTINO and J. ZHANG. -- A Markov Random Field modelbased approach to image interpretation. -- IEEE Trans. Pattern Anal. Machine Intell., Vol 14, No 6: pages 606--615, June 1992.

.... can be specified directly via Equation 3, i.e. via the specification of conditional probability tables or functions (Spiegelhalter et al. 1991) In contrast, UPINs must be specified in terms of clique functions (as in Equation 1) which may not be as easy to work with (cf. Geman and Geman (1984) Modestino and Zhang (1992) and Vandermeulen et al. 1994) for examples of ad hoc design of clique functions in image analysis) UPINs are more frequently used in problems such as image analysis and statistical physics where associations are thought to be correlational rather than causal. 3.4 From DPINs to (Decomposable) ....

Modestino, J. and Zhang, J. 1992. A Markov random field model-based approach to image segmentation. IEEE Trans. Patt. Anal. Mach. Int. 14(6), 606--615.

....(ii) using a predefined threshold to merge all segments whose area is less than the prespecified minimum area. This refinement reduces the number of segments in the k means segmented image. The problem of image interpretation is formulated in a MRF framework along the lines of Modestino and Zhang [10]. The optimal interpretation labels I are obtained by solving the MAP estimation problem I = arg max I2fLg n P[IjC(Y) K] where, K is the knowledge associated with the 3D scene, C(Y) are the core variables associated with the measurements made on the segmented image, n corresponds to the ....

J. A. Modestino and J. Zhang, "A Markov Random Field model based approach to image interpretation", IEEE Tran. on Pattern Analysis and Machine Intelligence, pp. 606--615, 1992.

.... image analysis and computer vision such as image restoration, 5, 11, 19] edge detection, 18] image segmentation, 13, 18] multisource image analysis, 24] computed tomography, 17] surface reconstruction, 13, 36] stereovision [2] motion analysis [9, 10, 23, 40] or scene interpretation, [37]. The mathematical framework is a statistical one: entities of interest in a given task are described by statistical models (Markov Random Fields) and bayesian estimation theory is used to extract the relevant information from the observed images. By defining comprehensive global statistical ....

J.W. MODESTINO and J. ZHANG. -- A Markov Random Field model-based approach to image interpretation. -- In IEEE Conf. Comp. Vision Pattern Rec., pages 458--465, June 1989.

....a single feedback process that incorporates contextual knowledge and they use genetic algorithm to produce an optimal image interpretation. Of late, Markov Random Field (MRF) models are being used for image interpretation with the view to make the interpretation systematic and domain independent [6, 7, 8]. Kim and Yang [9] integrate segmentation and interpretation to form a combined weighted energy function, the segmentation block is weighted high initially and as the algorithm iterates the weights shifts to the interpretation block. In this paper we propose a scheme for joint segmentation and ....

....to produce a crude segmented image. The segmented image is refined using the difference image (D Omega Gamma1 Y;HL , D Omega Gamma1 Y;LH , D Omega Gamma1 Y;HH ) as described in Section 3. The segmented image is interpreted in a MRF framework in a way analogous to Modestino and Zhang [6] except that we have a label corresponding to no interpretation as a possible label. The no interpretation labels are used as an aid to refine the segmented image before further interpretation is carried out. This process of interpretation, merging to produce better segmented image and ....

[Article contains additional citation context not shown here]

J. A. Modestino and J. Zhang. "A Markov Random Field Model Based Approach to Image Interpretation ". IEEE Trans. Patt. Anal. Machine Intell., pages 606--615, 1992.

....and the imaging process. Knowing the structures of objects and relations, means we can reduce the number of combinations to be produced in order to obtain the interpretation [7] Recent papers present many different types of interpretation systems: in [16, 5, 18] authors use expert systems, in [11] Markov Random Fields are used. Early work in knowledge based image interpretation has been summarized in [14, 1, 15] Methods for translating scene knowledge and constraints on the grouping and hypothesis generation process are crucial to limiting the number of either top down or bottom up ....

J.W. Modestino, J. Zhang, "A markov random field model-based approach to image interpretation", IEEE Pattern Analysis and Machine Intelligence, vol.14, no.6, June 1992, pp.606-615.

.... be specified directly via Equation 3, i.e. via the specification of con9 ditional probability tables or functions (Spiegelhalter et al. 1991) In contrast, UPINs must be specified in terms of clique functions (as in Equation 1) which may not be as easy to work with (cf. Geman and Geman (1984) Modestino and Zhang (1992) and Vandermeulen et al. 1994) for examples of ad hoc design of clique functions in image analysis) UPINs are more frequently used in problems such as image analysis and statistical physics where associations are thought to be correlational rather than causal. 3.4 From DPINs to (Decomposable) ....

Modestino, J. and Zhang, J. 1992. A Markov random field model-based approach to image segmentation. IEEE Trans. Patt. Anal. Mach. Int. 14(6), 606--615.

No context found.

J. W. Modestino and J. Zhang, "A Markov random field model-based approach to image interpretation," IEEE Trans. Pattern Anal. Machine Intell., vol. 14, pp. 606-- 615, June 1992.

No context found.

J. W. Modestino and J. Zhang, "A Markov random field model--based approach to image interpretation," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 14, pp. 606--615, June 1992.

No context found.

Modestino, J. and Zhang, J. 1992. A Markov random field model-based approach to image segmentation. IEEE Trans. Patt. Anal. Mach. Int. 14(6), 606--615.

No context found.

J.W. Modestino and J. Zhang, "A Markov random field model-based approach to image interpretation ". In Markov Random Fields: Theory and Applications, edited by R. Chellappa and A. Jain, pp. 369-408, Academic Press, 1993.

No context found.

J. W. Modestino and J. Zhang, "A Markov random field model-based approach to image interpretation," in Markov Random Fields: Theory and Applications, R. Chellappa and A. Jain, Eds. New York: Academic, 1993, pp. 369--408.

No context found.

J.W. Modestino and J. Zhang, "A Markov Random Field ModelBased Approach to Image Interpretation," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 14, no. 6, pp. 606--615, 1992.

No context found.

J. W. Modestino and J. Zhang, "A Markov random field model-based approach to image interpretation," IEEE Trans. Patt. Anal. Mach. Intell. 14, 6 (1992) 606--615.

No context found.

J.W. Modestino and J. Zhang. A Markov random field model-based approach to image interpretation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14:606--615, 1992.

No context found.

J.W. Modestino and J. Zhang. A Markov random field model-based approach to image interpretation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14:606--615, 1992.

First 50 documents

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