Markov random field image segmentation using cellular neural network

T. Szirányi, Josiane Zerubia

Research output: Contribution to journalArticle

23 Citations (Scopus)

Abstract

Markovian approaches to early vision processes need a huge amount of computing power. These algorithms can usually be implemented on parallel computing structures. With the Cellular Neural Networks (CNN), a new image processing tool is coming into consideration. Its VLSI implementation takes place on a single analog chip containing several thousands of cells. Herein we use the CNN UM architecture for statistical image segmentation. The Modified Metropolis Dynamics (MMD) method can be implemented into the raw analog architecture of the CNN. We are able to implement a (pseudo) random field generator using one layer (one memory/cell) of the CNN. We can introduce the whole pseudostochastic segmentation process in the CNN architecture using 8 memories/cell. We use simple arithmetic functions (addition, multiplication), equalitytest between neighboring pixels and very simple nonlinear output functions (step, jigsaw). With this architecture, a real VLSI CNN chip can execute a pseudostochastic relaxation algorithm of about 100 iterations in about 1 ms. In the proposed solution the segmentation is unsupervised. We have developed a pixellevel statistical estimation model. The CNN turns the original image into a smooth one. Then we have two graylevel values for every pixel: the original and the smoothed one. These two values are used for estimating the probability distribution of region label at a given pixel. Using the conventional firstorder Markov Random Field (MRF) model, some misclassification errors remained at the region boundaries, because of the estimation difficulties in case of low SNR. By using a greater neighborhood, this problem has been avoided. In our CNN experiments, we used a simulation system with a fixedpoint integer precision of 16 bits. Our results show that even in the case of the very constrained conditions of valuerepresentations (the interval is (64, +64), the accuracy is 0.002) can result in an effective and acceptable segmentation.

Original languageEnglish
Pages (from-to)8689
Number of pages1
JournalIEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications
Volume44
Issue number1
Publication statusPublished - 1997

Fingerprint

Cellular neural networks
Image segmentation
Pixels
Network architecture
Data storage equipment
Parallel processing systems
Probability distributions
Labels
Image processing

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

Markov random field image segmentation using cellular neural network. / Szirányi, T.; Zerubia, Josiane.

In: IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, Vol. 44, No. 1, 1997, p. 8689.

Research output: Contribution to journalArticle

@article{9af925fa31d44d518da0cb3fa0bfbae4,
title = "Markov random field image segmentation using cellular neural network",
abstract = "Markovian approaches to early vision processes need a huge amount of computing power. These algorithms can usually be implemented on parallel computing structures. With the Cellular Neural Networks (CNN), a new image processing tool is coming into consideration. Its VLSI implementation takes place on a single analog chip containing several thousands of cells. Herein we use the CNN UM architecture for statistical image segmentation. The Modified Metropolis Dynamics (MMD) method can be implemented into the raw analog architecture of the CNN. We are able to implement a (pseudo) random field generator using one layer (one memory/cell) of the CNN. We can introduce the whole pseudostochastic segmentation process in the CNN architecture using 8 memories/cell. We use simple arithmetic functions (addition, multiplication), equalitytest between neighboring pixels and very simple nonlinear output functions (step, jigsaw). With this architecture, a real VLSI CNN chip can execute a pseudostochastic relaxation algorithm of about 100 iterations in about 1 ms. In the proposed solution the segmentation is unsupervised. We have developed a pixellevel statistical estimation model. The CNN turns the original image into a smooth one. Then we have two graylevel values for every pixel: the original and the smoothed one. These two values are used for estimating the probability distribution of region label at a given pixel. Using the conventional firstorder Markov Random Field (MRF) model, some misclassification errors remained at the region boundaries, because of the estimation difficulties in case of low SNR. By using a greater neighborhood, this problem has been avoided. In our CNN experiments, we used a simulation system with a fixedpoint integer precision of 16 bits. Our results show that even in the case of the very constrained conditions of valuerepresentations (the interval is (64, +64), the accuracy is 0.002) can result in an effective and acceptable segmentation.",
author = "T. Szir{\'a}nyi and Josiane Zerubia",
year = "1997",
language = "English",
volume = "44",
pages = "8689",
journal = "IEEE Transactions on Circuits and Systems II: Express Briefs",
issn = "1057-7122",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "1",

}

TY - JOUR

T1 - Markov random field image segmentation using cellular neural network

AU - Szirányi, T.

AU - Zerubia, Josiane

PY - 1997

Y1 - 1997

N2 - Markovian approaches to early vision processes need a huge amount of computing power. These algorithms can usually be implemented on parallel computing structures. With the Cellular Neural Networks (CNN), a new image processing tool is coming into consideration. Its VLSI implementation takes place on a single analog chip containing several thousands of cells. Herein we use the CNN UM architecture for statistical image segmentation. The Modified Metropolis Dynamics (MMD) method can be implemented into the raw analog architecture of the CNN. We are able to implement a (pseudo) random field generator using one layer (one memory/cell) of the CNN. We can introduce the whole pseudostochastic segmentation process in the CNN architecture using 8 memories/cell. We use simple arithmetic functions (addition, multiplication), equalitytest between neighboring pixels and very simple nonlinear output functions (step, jigsaw). With this architecture, a real VLSI CNN chip can execute a pseudostochastic relaxation algorithm of about 100 iterations in about 1 ms. In the proposed solution the segmentation is unsupervised. We have developed a pixellevel statistical estimation model. The CNN turns the original image into a smooth one. Then we have two graylevel values for every pixel: the original and the smoothed one. These two values are used for estimating the probability distribution of region label at a given pixel. Using the conventional firstorder Markov Random Field (MRF) model, some misclassification errors remained at the region boundaries, because of the estimation difficulties in case of low SNR. By using a greater neighborhood, this problem has been avoided. In our CNN experiments, we used a simulation system with a fixedpoint integer precision of 16 bits. Our results show that even in the case of the very constrained conditions of valuerepresentations (the interval is (64, +64), the accuracy is 0.002) can result in an effective and acceptable segmentation.

AB - Markovian approaches to early vision processes need a huge amount of computing power. These algorithms can usually be implemented on parallel computing structures. With the Cellular Neural Networks (CNN), a new image processing tool is coming into consideration. Its VLSI implementation takes place on a single analog chip containing several thousands of cells. Herein we use the CNN UM architecture for statistical image segmentation. The Modified Metropolis Dynamics (MMD) method can be implemented into the raw analog architecture of the CNN. We are able to implement a (pseudo) random field generator using one layer (one memory/cell) of the CNN. We can introduce the whole pseudostochastic segmentation process in the CNN architecture using 8 memories/cell. We use simple arithmetic functions (addition, multiplication), equalitytest between neighboring pixels and very simple nonlinear output functions (step, jigsaw). With this architecture, a real VLSI CNN chip can execute a pseudostochastic relaxation algorithm of about 100 iterations in about 1 ms. In the proposed solution the segmentation is unsupervised. We have developed a pixellevel statistical estimation model. The CNN turns the original image into a smooth one. Then we have two graylevel values for every pixel: the original and the smoothed one. These two values are used for estimating the probability distribution of region label at a given pixel. Using the conventional firstorder Markov Random Field (MRF) model, some misclassification errors remained at the region boundaries, because of the estimation difficulties in case of low SNR. By using a greater neighborhood, this problem has been avoided. In our CNN experiments, we used a simulation system with a fixedpoint integer precision of 16 bits. Our results show that even in the case of the very constrained conditions of valuerepresentations (the interval is (64, +64), the accuracy is 0.002) can result in an effective and acceptable segmentation.

UR - http://www.scopus.com/inward/record.url?scp=84866217698&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84866217698&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84866217698

VL - 44

SP - 8689

JO - IEEE Transactions on Circuits and Systems II: Express Briefs

JF - IEEE Transactions on Circuits and Systems II: Express Briefs

SN - 1057-7122

IS - 1

ER -