J.-P. Wang, Stochastic relaxation on partitions with connected components and its application to image segmentation, IEEE Transactions on Pattern Analysis and Machine Intelligence 20 (6) (1998) 619 -- 626.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....before performing high level tasks such as object recognition and content based image retrieval. Because of its uniqueness and importance in many computer vision applications, a variety of techniques have been proposed in the past research. Recent work includes stochastic model based approaches [2, 10], curve evolution [6] energy diffusion [7] region growing [1, 5] and graph partitioning [9, 12] Quantitative evaluation methods have also been suggested [3, 8] As stated in [5] texture is the main difficulty faced to a segmentation method. For the images that contain only homogeneous color ....

J.-P. Wang, "Stochastic relaxation on partitions with connected components and its application to image segmentation", PAMI, vol. 20, no.6, pp. 619-36, 1998.

....the whole image is examined at the same time. This approach generally assumes that the image contains parts of objects and their boundaries are the set of curves that minimizes a global energy function. Some energy models are based on a discrete model of the image, such as Markov random fields ([18, 27, 47] whereas variational models are based on a continuous model of the image [5, 35, 34, 33, 44, 21] Leclerc [27] proposed a partition process based on a Minimum Description Length (MDL) representation of both the intensity variation within region and the enclosed boundary. Finally, Blake and ....

....for Markov random fields, but given by the components which are adjacent to the pixel. These models are especially applicable for images where a relatively few number of gray levels occur, and where some prior knowledge is available about size and shape characteristics for the connected components [47]. The Markov connected component fields possess certain appealing Markov properties which have been established in [31] Here we considered a Markov connected component field whose probability density function is proportional to exp Gamma ffl 1 P ffl 2 i j ffl 3 Ep( Omega ....

J.P. Wang. Stochastic relaxation on partitions with connected components and its application to image segmentation. IEEE Trans. Patt. Anal. and Mach. Int., 20(6):619--636, 1998.

....However, the problem of unsupervised segmentation is ill defined because seman tic objects do not usually correspond to homogeneous spatio temporal regions in color, texture, or motion. Some of the recent work in image segmentation include stochastic model based approaches [1] 6] 13] 17] [24], 25] morphological watershed based region growing [18] energy diffusion [ 14] and graph partitioning [20] The work on video segmentation include motion based segmentation [3] 19] 21] 23] spatial segmentation and motion tracking [8] 22] moving objects extraction [12] 15] and ....

J.-P. Wang, "Stochastic relaxation on partitions with connected components and its application to image segmentation," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 20, no.6, p. 619-36, 1998.

....properties of an image and to develop efficient algorithms for computing segmentations. There is a wide range of image segmentation techniques in the literature. Many of them rely on the design and minimization of an energy function which captures the interaction between models and image data [10, 2, 21, 15, 27]. Conventional segmentation techniques generally fall into two distinct classes, being either boundary based or region based. The former class looks at the image discontinuities near objects boundaries, while the latter examines the homogeneity of spatially localized features inside objects ....

....are the set of curves that minimizes a global energy function. Past approaches have centered on formulating the problem as the minimization of a functional involving the image intensity and edge functions. Some energy models are based on a discrete model of the image, such as Markov random fields [10, 15, 27] or a Minimum Description Length (MDL) representation [15, 8] whereas variational models are based on a continuous model of the image [2, 21, 20, 19, 26] More recently, Zhu attempted to unify snakes, region growing and energy Bayes MDL within a general framework [29] Finally, Blake and ....

[Article contains additional citation context not shown here]

J. Wang. Stochastic relaxation on partitions with connected components and its application to image segmentation. IEEE Trans. Patt. Anal. and Mach. Int., 20(6):619--636, 1998.

....explaining the observations is the best [11, 21, 24, 39] Finding the globally consistent, MDL image labeling is impractical in general due to the computational complexity of global optimization algorithms. This has led to the use of parallel algorithms [11, 24] or approximation algorithms [5, 8, 15, 21, 29, 32, 37, 39]. 2 Overview of Approach The above mentioned work leads to the development of our approach. Deformable shape templates are used to partition the image into a globally consistent interpretation, determined in part by the MDL principle. The formulation can be used to group image regions based on ....

....of labels in the image. Such information is not always available for general imagery. After defining the criterion function for labeling, the next problem is computing the solution to the optimization problem. A number of proposed approaches employ simulated or deterministic annealing [5, 32, 15, 29, 37] (for a comparison see [25] Chou and Brown [8] used highest confidence first (HCF) to infer a unique labeling from the posteriori distribution that is consistent with both the prior knowledge and evidence. Their method is analgous to deterministic annealing, but computation is more efficient. A ....

J. Wang. Stochastic relaxation on partitions with connected components and its application to image segmentation. PAMI, 20(6):619--636, 1998.

....in classification, neural networks have been applied to perform texture segmentation. Chellappa et al. 3] employed a Hopfield type network to maximize aposteriori probability (MAP) for texture segmentation. Unlike the approaches employing known filter forms, such as Gabor filters [1] 10] [23], 9] Jain and Karu trained a multilayer perceptron (MLP) to obtain texture specific filters and to perform segmentation [11] The biological relevance of these studies is limited as are the results on real texture. Eom [6] segmented texture images using a neural network whereby a two dimensional ....

....complexity, for a network of the size 80 80, our network simulation takes typically a minute of CPU time on an HP workstation. This computational performance is comparable with the steepest descent and Monte Carlo algorithms used in those MAP based approaches that reported computing times [23]. To summarize, after carefully evaluating our method against existing methods in terms of both methodology and performance, we are led to the conclusion that our method offers unique computational advantages, in addition to the fact that our approach has a strong biological link. V. CONCLUSION ....

J.-P. Wang, "Stochastic relaxation on partitions with connected components and its applications to image segmentation," IEEE Trans. Pattern Anal. Machine Intell., vol. 20, pp. 619--636, 1998.

No context found.

J.-P. Wang, Stochastic relaxation on partitions with connected components and its application to image segmentation, IEEE Transactions on Pattern Analysis and Machine Intelligence 20 (6) (1998) 619 -- 626.

No context found.

J. P. Wang, \Stochastic Relaxation on Partitions with Connected Components and Its Applications to Image Segmentation," IEEE Trans. PAMI, vol. 20, no. 6, pp. 619-636, June 1998.

No context found.

J-P.Wang, "Stochastic relaxation on partitions with connected components and its application to image segmentation", IEEE Trans. PAMI, Vol. 20, No. 6, pp 619-636, 1996.

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