D. A. Bader, J. JaJa, and R. Chellappa, "Scalable data parallel algorithms for texture synthesis using Gibbs random fields," IEEE Transactions on Image Processing, vol. 4, no. 10, pp. 1456--1460, 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Paget, D. Longsta#, and B. Lovell, Texture classification using nonparametric Markov random fields, in Proceedings of the 1997 13th International Conference on Digital Signal Processing, vol. 1, Hellas, Santorini, Greece) pp. 67 70, IEEE Signal Processing Society, July 1997. Invited Paper. [6] R. Paget and D. Longsta#, Extracting the cliques from a neighbourhood system, IEE Proceedings Vision, Image and Signal Processing, vol. 144, pp. 168 170, June 1997. 7] R. Paget and D. Longsta#, A nonparametric multiscale Markov random field model for synthesising natural textures, in ....

.... Conventional texture models, like the auto models [17] autoregressive (AR) models [37] moving average (MA) models [95] or combination of both (ARMA) models [115] have not been found to provide a basis for realistically synthesising natural textures [95] Although Bader, JaJa and Chellappa [6] did achieve modest results with a GMRF model. However recent advances in texture synthesis have produced models that are capable of synthesising natural textures [28] textures that contain both structural and statistical elements. These models are based on the stochastic modelling of various ....

[Article contains additional citation context not shown here]

D. A. Bader, J. JaJa, and R. Chellappa, "Scalable data parallel algorithms for texture synthesis using Gibbs random fields," IEEE Transactions on Image Processing, vol. 4, no. 10, pp. 1456--1460, 1995.

....Processing Society, July 1997. Invited Paper. R. Paget and D. Longstaff, Extracting the cliques from a neighbourhood sys 170, June 1997. R. Paget and D. Longstaff, A nonparametric multiscale Markov random field pp. 744 747, ISSPA 96, August 1996. xxiv PUBLICATIONS ARISING xxv Is] [9] [10] 11] 12] 13] 14] 15] R. Paget and D. Longstaff, Non parametric Markov random field texture synthesis. CSSIP Tekfest Centre for Sensor Signal and Information Pro cessing, February 1996. R. Paget and D. Longstaff, Texture synthesis via a nonparametric Markov random field, in ....

.... Conventional texture models, like the auto models [22] autoregressive (AR) models [51] moving average (MA) models [142] or combination of both (ARMA) models [182] have not been found to provide a basis for realistically synthesising natural textures [142] Although Bader, J J and Chellappa [9] did achieve modest results with a GMRF model. However recent advances in texture synthesis have pro duced models that are capable of synthesising natural textures [39] textures that contain both structural and statistical elements. These models are based on the was textural information in the ....

[Article contains additional citation context not shown here]

D. A. Bader, J. JJ, and R. Chellappa, "Scalable data parallel algorithms for texture synthesis using Gibbs random fields," IEEE Transactions on Image Processing, vol. 4, no. 10, pp. 1456 1460, 1995.

....than the monogrid model to achieve a good result. To decide which one to use in a VLSI chip environment depends on the available technological potential. Preprocessing There are several attempt to introduce dierent image processing methods in the MRF framework: texture analysis and synthesis [3], motion detection [7] or restoration from blurred and noisy images [18] These combined methods are usually very specic for the given task. Sometimes, the composite algorithm is eective only for binary imaging eects, such as simple motiondetection or deconvolution to get binary image segments ....

Bader, D.A. Jala, J., & Chellappa, R. (1995) Scalable data parallel algorithms for texture synthesis using Gibbs random elds. IEEE Tr. Image Processing, 4: 14561460.

No context found.

D. A. Bader, J. JaJa, and R. Chellappa. Scalable Data Parallel Algorithms for Texture Synthesis Using Gibbs Random Fields. IEEE Trans. Image Processing, 4(10):1456--1460, 1995.

....image size, and neighborhood models can be found in Tables 8, 9, 10, and 11, for a 4th order model on this selection of real world textured images, and in Tables 12, 13, 14, and 15, for a higher order model on the same set of images. Similarly, CM 2 timings for these estimates can be found in [2]. Tables 8 15 are given in Appendix C. 21 5 Texture Compression We implement an algorithm for compressing an image of a GMRF texture to approximately 1 bit pixel from the original 8 bits pixel image. The procedure is to find the MLE of the given image, e.g. this results in a total of eleven ....

D. A. Bader, J. J'aJ'a, and R. Chellappa. Scalable Data Parallel Algorithms for Texture Synthesis and Compression Using Gibbs Random Fields. Technical Report CS-TR-3123 and UMIACS-TR-93-80, UMIACS and Electrical Engineering, University of Maryland, College Park, MD, August 1993.

No context found.

D. A. Bader, J. JaJa, and R. Chellappa. Scalable Data Parallel Algorithms for Texture Synthesis Using Gibbs Random Fields. IEEE Trans. Image Processing, 4(10):1456--1460, 1995.

....each pixel location, scalable, and simple. An example of a binary synthetic texture generated by the Gibbs Sampler is given in Figure 4. Table 1 shows the timings of a binary Gibbs sampler for model orders 1, 2, and 4, on the CM 5. More extensive tables for both the CM 2 and CM 5 can be found in [1]. 3 2.2 Gaussian Markov Random Field Sampler In this section, we consider the class of 2 D non causal models called the Gaussian Markov random field (GMRF) models described in ( 3] 5] and [9] Pixel grey levels have joint Gaussian distributions and correlations controlled by a number of ....

....of the M 2 values in the logarithm term for F is positive. Finally, an optimality test is performed. We set Theta k 1 = Theta k #, and if Theta k 1 is sufficiently close to Theta k , the procedure terminates. We give the first and second derivatives of F with respect to Theta k and in [1]. 6 For a rapid convergence of the Newton Raphson method, it must be initialized with a good estimate of parameters close to the global maximum. We use the least squares estimate given in Subsection 3.1 as Theta 0 , the starting value of the parameters. In Figure 5, we show the synthesis using ....

[Article contains additional citation context not shown here]

D. A. Bader, J. J'aJ'a, and R. Chellappa. Scalable Data Parallel Algorithms for Texture Synthesis and Compression Using Gibbs Random Fields. Technical Report CS-TR-3123 and UMIACS-TR-93-80, UMIACS and Electrical Engineering, University of Maryland, College Park, MD, August 1993. 7

....data on the nodes have been shown to have little contention in the fat tree network. This makes the network very efficient for regular communications patterns commonly used in massively parallel processing, since highly parallel code utilizes permutations when performing data parallel grid shifts ([1]) or for common mathematical operations such as the Cooley Tukey Fast Fourier Transform ( 4] 5] The fat tree interconnection network is a packet switching network. Each processing node is responsible for splitting messages into data packets, and injecting and receiving the packets from the ....

.... number of keys in a packet to send double bufferlenMB; CMMDfsetiomode(stdout, CMMDindependent) pid = CMMDselfaddress( This is our PE ID numnodes = CMMDpartitionsize( This is the size of our partition PACKETSIZE = PACKETSIZEDEFAULT; dp = DPDEFAULT; if (argc = 2) dp = atoi(argv[1]) if (argc = 3) PACKETSIZE = atoi(argv[2] if (pid = PIDTIMER) printf( Partition: 2d Packetsize: 4d Shift: 2d n , numnodes, PACKETSIZE, dp) bufferlenMB = MSGLEN PACKETSIZE pow(2.0,20.0) 18 send packet sendpacket = int )malloc(sizeof(int) PACKETSIZE) checkptr(sendpacket) ....

[Article contains additional citation context not shown here]

David A. Bader, Joseph J'aJ'a, and Rama Chellappa. Scalable Data Parallel Algorithms for Texture Synthesis and Compression Using Gibbs Random Fields. Technical Report CS-TR-3123 and UMIACS-TR-93-80, UMIACS and Electrical Engineering, University of Maryland, College Park, MD, August 1993.

....image size, and neighborhood models can be found in Tables 8, 9, 10, and 11, for a 4th order model on this selection of real world textured images, and in Tables 12, 13, 14, and 15, for a higher order model on the same set of images. Similarly, CM 2 timings for these estimates can be found in [2]. Tables 8 15 are given in Appendix C. 5 Texture Compression We implement an algorithm for compressing an image of a GMRF texture to approximately 1 bit pixel from the original 8 bits pixel image. The procedure is to find the MLE of the given image, e.g. this results in a total of eleven ....

D. A. Bader, J. J'aJ'a, and R. Chellappa. Scalable Data Parallel Algorithms for Texture Synthesis and Compression Using Gibbs Random Fields. Technical Report CS-TR-3123 and UMIACS-TR-93-80, UMIACS and Electrical Engineering, University of Maryland, College Park, MD, August 1993.

No context found.

D. A. Bader, J. JaJa, and R. Chellappa, "Scalable data parallel algorithms for texture synthesis using Gibbs random fields," IEEE Transactions on Image Processing, vol. 4, no. 10, pp. 1456--1460, 1995.

No context found.

D. A. Bader, J. JaJa, and R. Chellappa, "Scalable data parallel algorithms for texture synthesis using Gibbs random fields," IEEE Transactions on Image Processing, vol. 4, no. 10, pp. 1456--1460, 1995.

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