Academia.eduAcademia.edu

Outline

Title
Abstract
Figures
Introduction
Proposed Method
Experimental Results
Conclusions
References
All Topics
Computer Science
Information Systems

A reinforcement agent for threshold fusion

Profile image of Maryam ShokriMaryam Shokri

2008, Applied Soft Computing

https://doi.org/10.1016/J.ASOC.2006.12.003
visibility

…

description

8 pages

link

1 file

Abstract

Finding an optimal threshold in order to segment digital images is a difficult task in image processing. Numerous approaches to image thresholding already exist in the literature. In this work, a reinforced threshold fusion for image binarization is introduced which aggregates existing thresholding techniques. The reinforcement agent learns the optimal weights for different thresholds and segments the image globally. A fuzzy reward function is employed to measure object similarities between the binarized image and the original gray-level image, and provide feedback to the agent. The experiments show that promising improvement can be obtained. Three well-established thresholding techniques are combined by the reinforcement agent and the results are compared using error measurements based on ground-truth images.

Figures (10)
Fig. 1. Basic components of reinforcement learning.
Fig. 1. Basic components of reinforcement learning.
Q-learning algorithm [23] Table 2
Q-learning algorithm [23] Table 2
Fig. 2. Overall structure of the reinforced threshold fusion for image binariza- tion. The selected thresholding techniques deliver their thresholds to the agent. Q-learning [26](Table 1) is applied as the RL algorithm. It is the most widely known and employed reinforcement learning algorithm. We mainly selected it because of its simple implementation and reportedly good performance. An optimal policy must be defined in order to make action selection possible. It enables the agent to map the current states to proper actions. We use the é-greedy policy
Fig. 2. Overall structure of the reinforced threshold fusion for image binariza- tion. The selected thresholding techniques deliver their thresholds to the agent. Q-learning [26](Table 1) is applied as the RL algorithm. It is the most widely known and employed reinforcement learning algorithm. We mainly selected it because of its simple implementation and reportedly good performance. An optimal policy must be defined in order to make action selection possible. It enables the agent to map the current states to proper actions. We use the é-greedy policy
Definitions for weight adjustment factor A (constants and intervals are selected heuristically)
Definitions for weight adjustment factor A (constants and intervals are selected heuristically)
Fig. 4. From left to right: original images; corresponding ground truth images; results of Otsu method; results of Kittler method; results of Kapur method; results of proposed technique.
Fig. 4. From left to right: original images; corresponding ground truth images; results of Otsu method; results of Kittler method; results of Kapur method; results of proposed technique.
The error measure x (Eq. (7)) for the selected algorithms and their RL fusion
The error measure x (Eq. (7)) for the selected algorithms and their RL fusion
Overall performance (O,) of thresholding techniques based on ranking mea- surement
Overall performance (O,) of thresholding techniques based on ranking mea- surement
Fig. 5. Typical curves for accumulated rewards (plots of average reward vs. number of iterations). The average number of iterations for convergence of the algorithm is ~ 63 for illustrated cases. The typical curves for accumulated rewards (plots of average reward versus number of iterations) are presented in Fig. 5. As it can be seen from Table 6, the proposed technique reaches the highest ranking mainly because it has the lowest error in 8 cases and, in contrast to the fused methods, has not been the last choice for any of the tested images.
Fig. 5. Typical curves for accumulated rewards (plots of average reward vs. number of iterations). The average number of iterations for convergence of the algorithm is ~ 63 for illustrated cases. The typical curves for accumulated rewards (plots of average reward versus number of iterations) are presented in Fig. 5. As it can be seen from Table 6, the proposed technique reaches the highest ranking mainly because it has the lowest error in 8 cases and, in contrast to the fused methods, has not been the last choice for any of the tested images.
Average computing times of Otsu, Kittler, Kapur, and proposed technique for presented results (all methods implemented in MATLAB)
Average computing times of Otsu, Kittler, Kapur, and proposed technique for presented results (all methods implemented in MATLAB)

Related papers

Supervised grayscale thresholding based on transition regions
Qingmao Hu

Image and Vision Computing, 2008

A new thresholding framework is proposed which is transition region based, and consists of deriving the transition region with the help of supervision and calculating the threshold from the transition region. Four ways of supervision are studied: picking up an object and a background pixel, from other clustering or segmentation results, based on sample statistics, and exploration of background proportions. The approach has been validated both quantitatively and qualitatively. It is found that the proposed approach: (1) is more robust, consistent and reliable than the conventional transition-region-based thresholding methods; and (2) is easier to implement and has wider applicability than existing supervised thresholding methods. The approach is especially useful for segmenting difficult images with multiple objects and/or serious imaging artifacts.

View PDFchevron_right
Optimization Methods for Image Thresholding: A review
Azmi Alrawi

Maǧallaẗ ǧāmiʻaẗ al-anbār li-l-ʻulūm al-ṣirfaẗ, 2023

Setting a border with the proper gray level in processing images to separate objects from their backgrounds is crucial. One of the simplest and most popular methods of segmenting pictures is histogram-based thresholding. Thresholding is a common technique for image segmentation because of its simplicity. Thresholding is used to separate the Background of the image from the Foreground. There are many methods of thresholding. This paper aims to review many previous studies and mention the types of thresholding. It includes two types: the global and local thresholding methods and each type include a group of methods. The global thresholding method includes (the Otsu method, Kapur's entropy method, Tsallis entropy method, Hysteresis method, and Fuzzy entropy method), and the local thresholding method includes (Ni-Black method and Bernsen method). The optimization algorithms(Genetic Algorithm, Particle Swarm Optimization, Bat Algorithm, Modified Grasshopper Optimization, Firefly Algorithm, Cuckoo Search, Tabu Search Algorithm, Simulated Annealing, and Jaya Algorithm) used along with thresholding methods are also illustrated.

View PDFchevron_right
An Adaptive Multi-Thresholding Technique for Binarization of Color Images
Imtiaz Hassan

Applying binarization technique on a colored image can yield an image that is quite different from the original one if multi-thresholding is used. The solution proposed is an adaptive multi-thresholding technique (AMTT) that describes an adaptive multi-thresholding technique (AMTT) for binarization of color images depending upon the nature of the image. The technique improves the perception of the binarized images based on color hue and hence the dissimilarity between original and binarized image is reduced. This technique compensates for differences in illumination and shade by including information content in the thresholding calculation. AMTT calculates information loss in the image with respect to human perception on the basis of color hue. Experimental results show that the proposed AMTT reduces the color content by keeping the information loss minimum.

View PDFchevron_right
An image segmentation algorithm in image processing based on threshold segmentation
Kamel Belloulata

Proceedings - International Conference on Signal Image Technologies and Internet Based Systems, SITIS 2007, 2007

Image segmentation is a key technology in image processing, and threshold segmentation is one of the methods used frequently. Aimed at that only one threshold or several thresholds are set in traditional threshold-based segmentation algorithm, it is difficult to extract the complex information in an image, a new segmentation algorithm that each pixel in the image has its own threshold is proposed. In this algorithm, the threshold of a pixel in an image is estimated by calculating the mean of the grayscale values of its neighbor pixels, and the square variance of the grayscale values of the neighbor pixels are also calculated as an additional judge condition, so that the result of the proposed algorithm is the edge of the image. In fact the proposed algorithm is equal to an edge detector in image processing. Experimental results demonstrate that the proposed algorithm could produce precise image edge, while it is reasonable to estimate the threshold of a pixel through the statistical information of its neighbor pixels.

View PDFchevron_right
A New Local Adaptive Thresholding Technique in Binarization
Romen Taiyenjam

2012

Image binarization is the process of separation of pixel values into two groups, white as background and black as foreground. Thresholding plays a major in binarization of images. Thresholding can be categorized into global thresholding and local thresholding. In images with uniform contrast distribution of background and foreground like document images, global thresholding is more appropriate. In degraded document images, where considerable background noise or variation in contrast and illumination exists, there exists many pixels that cannot be easily classified as foreground or background. In such cases, binarization with local thresholding is more appropriate. This paper describes a locally adaptive thresholding technique that removes background by using local mean and mean deviation. Normally the local mean computational time depends on the window size. Our technique uses integral sum image as a prior processing to calculate local mean. It does not involve calculations of standard deviations as in other local adaptive techniques. This along with the fact that calculations of mean is independent of window size speed up the process as compared to other local thresholding techniques.

View PDFchevron_right
Effective Histogram Thresholding Techniques for Natural Images Using Segmentation
Dr. K.K. Thanammal

Journal of Image and Graphics, 2015

View PDFchevron_right
Image Thresholding using Histogram Fuzzy Approximation
Mohammad Al-Azawi

International Journal of Computer Applications 83(9):36-40, December 2013. Published by Foundation of Computer Science, New York, USA, 2013

Image segmentation is one of the most important techniques in image processing. It is widely used in different applications such as computer vision, digital pattern recognition, robot vision, etc. Histogram was the earliest feature that has been used for isolating objects from their background, it is widely applicable in different application in which one needs to divide the image into distinct regions like background and object. The thresholding technique is the most popular solution in which a value on the histogram is selected to separate the regions. This value, which is known as the threshold, should be specified in an appropriate way. One of the methods is by using the global minimum value of the histogram and divides the histogram into white and black (binary image). Due to the spatial and grey uncertainty and ambiguity, the extraction of the threshold value in a crispy way is not suitable always. To overcome such problems, the proposed method uses two membership functions to measure the whiteness and blackness of a member element. The pixel belonging to one of the region is dependent on the membership value it has according to the membership functions.

View PDFchevron_right
On creation of reference image for quantitative evaluation of image thresholding method
Nabendu Chaki

Communications in Computer and Information Science, 2011

A good reference image is important for relative performance analysis of different image thresholding techniques in a quantitative manner. There exist standard methods for building reference images for document image binarization. However, a gap is found for graphic images referencing. This paper offers six different techniques for building reference images. These may be used for comparing different image thresholding techniques. Experimental results illustrate the relative performance of five different image thresholding methods for the six reference image building methods on a set of ten images taken from the USC-SIPI database. The results would help picking up the right reference image for evaluating binarization techniques.

View PDFchevron_right
Survey over image thresholding techniques and quantitative performance evaluation
nk Rajgure

Journal of Electronic Imaging, 2004

In this study we conduct an exhaustive survey of image thresholding methods, categorize them and express their formulae under a uniform notation, and finally carry their performance comparison. The thresholding methods are categorized according to the information they are exploiting, such as: histogram shape, measurement space clustering, entropy, object attributes, spatial correlation and local gray-level surface. Forty selected thresholding methods from various categories are compared in the context of non-destructive testing applications as well as for document images. The comparison is based on the combined performance measures. We identify the thresholding algorithms that perform uniformly better over nondestructive testing and document image applications.

View PDFchevron_right
A Brief Review of Optimization of Methods for Image Segmentation by using Thresholding Techniques
Neha Baghel

International Journal for Research in Applied Science and Engineering Technology, 2020

Image segmentation is important part of computer vision and image processing. The applicability and diversity of image segmentation are increase day to day in various engineering and scientific filed for the purpose of data analysis and prediction of particular object in given image [10]. For the processing of image segmentation various technique are used some technique are based on histogram of image and some technique are based on image content such as color, texture and shape & size. In mid-decade used the concept of threshold based image segmentation technique. Threshold based image segmentation technique overcomes the limitation of pervious method of image segmentation. The threshold based image segmentation method performs in terms of local, global and adaptive image segmentation techniques. The process of local and global image segmentation technique differs only in the selection of parameter for the threshold. The selection of threshold value includes the process of image binarization. The local and global image segmentation technique is based on the method of iteration [11]. The process of image iteration cannot be always good for image similarity index for segmented area. In this dissertation the analysis of image segmentation technique is performed based on thresholding technique. For the evaluation of the algorithm performance execution time is used. For the validation of local and global algorithm some standard image dataset is used such as boat, cameraman and Barbara.

View PDFchevron_right

References (35)

  1. A. Ayesh, Emotionally motivated reinforcement learning based controller, in: Proceedings of the IEEE SMC, The Hague, The Netherlands, 2004.
  2. H.R. Berenji, Fuzzy Q-learning: a new approach for fuzzy dynamic programming problems, in: Proceedings of the Third IEEE International Conference on Fuzzy Systems, Orlando, FL, June 1994.
  3. D. Bhandari, N.R. Pal, D.D. Majumder, Fuzzy divergence, probability measure of fuzzy events and image thresholding, Pattern Recogn. Lett. 13 (1992) 857-867.
  4. K.R. Castleman, Digital Image Processing, Prentice Hall, Upper Saddle River, New Jersey, 1998, 07458.
  5. S. Gadanho, Reinforcement learning in autonomous robots: an empirical investigation of the role of emotions, Ph.D. Thesis, University of Edin- burgh, Edinburgh, 1999.
  6. P.Y. Glorennec, Fuzzy Q-learning and dynamical fuzzy Q-learning, in: Proceedings of the Third IEEE International Conference on Fuzzy Sys- tems, IEEE Press, Piscataway, NJ, 1994, pp. 474-479.
  7. P.Y. Glorennec, L. Jouffe, Fuzzy Q-learning, in: Proceedings of the Sixth International Conference on Fuzzy Systems, Barcelona, Spain, (1997), pp. 659-662.
  8. R.C. Gonzalez, R.E. Woods, Digital Image Processing, 2nd ed., Prentice Hall, New Jersey, 2002, 07458.
  9. W.D. Gregg, Analog and Digital Communication, John Wiley and Sons, Inc., USA, 1977.
  10. R.W. Hamming, Error detecting and error correcting codes, Bell System Tech. J. 29 (1950) 933-968.
  11. L.K. Huang, M.J.J. Wang, Image thresholding by minimizing the mea- sures of fuzziness, Pattern Recogn. 28 (1995) 41-51.
  12. L. Jouffe, Fuzzy inference system learning bay reinforcement methods, IEEE Trans. Syst. Man Cybern. 28 (1998) 338-355.
  13. L.P. Kaelbling, M.L. Littman, A.W. Moore, Reinforcement learning: a survey, J. Artif. Intell. Res. 4 (1996) 237-285.
  14. J.N. Kapur, P.K. Sahoo, A.K.C. Wong, A new method for gray-level picture thresholding using the entropy of the histogram, Comput. Vis. Graph. Image Process. 29 (1985) 273-285.
  15. J. Kittler, J. Illingworth, Minimum error thresholding, Pattern Recogn. 19 (1) (1986) 41-47.
  16. S.U. Lee, S.Y. Chung, R.H. Park, A Comparative performance study of several global thresholding techniques for segmentation, Comput. Vis. Graph. Image Process. 52 (1990) 171-190.
  17. N. Otsu, A threshold selection method from gray level histogram, IEEE Trans. Syst. Man Cybern. SMC-9 (1979) 62-66.
  18. S.J. Russell, P. Norvig, Artificial Intelligence: A Modern Approach, Prentice Hall, Englewood Cliffs, N.J., 1995.
  19. P.K. Sahoo, S. Soltani, A.K.C. Wong, Y.C. Chen, A survey of thresholding technique, Comput. Vis. Graph. Image Process. 41 (1988) 233-260.
  20. B. Sankur, M. Sezgin, Survey over image thresholding techniques and quantitative performance evaluation, J. Electron. Imaging 13 (1) (2004) 146-165.
  21. M. Shokri, H.R. Tizhoosh, Using reinforcement learning for image thresholding, Can. Conf. Electr. Comput. Eng. 1 (2003) 1231-1234.
  22. M. Shokri, H.R. Tizhoosh, Q(l)-based image thresholding, Can. Conf. Comput. Robot Vis. (2004) 504-508.
  23. R.S. Sutton, A.G. Barto, Reinforcement Learning: An Introduction, MIT Press, Cambridge, Mass, 1998.
  24. W. Tsai, Moment-preserving thresholding: a new approach, Comput. Vis. Graph. Image Process. 29 (1985) 377-393.
  25. M.A. Walker, An application of reinforcement learning to dialogue strategy selection in a spoken dialogue system for email, J. Artif. Intell. Res. 12 (2000) 387-416.
  26. C.J.C.H. Watkins, Learning from Delayed Rewards, Cambridge Uni- versity, Cambridge, 1989.
  27. C.J.H. Watkins, P. Dayan, Technical note, Q-learning, Mach. Learn. 8 (1992) 279-292.
  28. H. Yan, Unified formulation of a class of image thresholding techniques, Pattern Recogn. 29 (12) (1996) 2025-2032.
  29. P. Yin, Maximum entropy-based optimal threshold selection using deter- ministic reinforcement learning with controlled randomization, Signal Process. 82 (2002) 993-1006.
  30. J. Yoshimoto, Sh. Ishii, M. Sato, Application of reinforcement learning to balancing of acrobat, IEEE Int. Conf. Syst. Man Cybern. (2003) 516-521.
  31. W. Zhang, T.G. Dietterich, Value function approximations and job-shop scheduling, Workshop on Value Function Approximation in Reinforce- ment Learning at ICML-95, 1995, submitted.
  32. S. Ali, K.A. Smith, On learning algorithm selection for classification, Appl. Soft Comput. J. 6 (2) (2006) 119-138.
  33. I. Ko ´kai, A. Lo ¨rincz, Fast adapting value estimation-based hybrid archi- tecture for searching the world-wide web, Appl. Soft Comput. J. 2 (1) (2002) 11-23.
  34. M. Shokri, H.R. Tizhoosh, M. Kamel, The outline of reinforcement learning technique for e-learning applications, in: S. Pierre (Ed.), E- Learning Networked Environments and Architectures: A Knowledge Processing Perspective, Springer Book Series, 2007.
  35. B. Bhanu, Y. Lin, Object detection in multi-modal images using genetic programming, Appl. Soft Comput. J. 4 (2) (2004) 175-201.

Related papers

A New Approach of Gray Images Binarization with Threshold Methods
Daniela Hossu

arXiv (Cornell University), 2015

The paper presents some aspects of the (gray level) image binarization methods used in artificial vision systems. It is introduced a new approach of gray level image binarization for artificial vision systems dedicated to industrial automationtemporal thresholding. In the first part of the paper are extracted some limitations of using the global optimum thresholding in gray level image binarization. In the second part of this paper are presented some aspects of the dynamic optimum thresholding method for gray level image binarization. Starting from classic methods of global and dynamic optimal thresholding of the gray level images in the next section are introduced the concepts of temporal histogram and temporal thresholding. In the final section are presented some practical aspects of the temporal thresholding method in artificial vision applications form the moving scene in robotic automation class; pointing out the influence of the acquisition frequency on the methods results.

View PDFchevron_right
A New Approach of Gray Images Binarization for Artificial Vision Systems with Threshold Methods
Daniela Hossu

Proceedings of the Fifth International Conference on Informatics in Control, Automation and Robotics Service, 2008

View PDFchevron_right
From Segmentation to Binarization of Gray-Level Images
Gabriella Sanniti di Baja

Journal of Pattern Recognition Research, 2008

For some gray-level images, the boundary between the foreground and the background is perceived in correspondence with the locally maximal changes in gray-level through the image. In this framework, this paper proposes a method to extract the objects of interest from an image and, hence, to distinguish the foreground from the background, starting from a partition of the image obtained by means of watershed transformation. The regions that are assigned to the foreground are also hierarchically ranked, depending on their perceptual relevance, so that different representations of the image are possible.

View PDFchevron_right
A survey of thresholding techniques
Prasanna Sahoo

Graphical Models /graphical Models and Image Processing /computer Vision, Graphics, and Image Processing, 1988

In digital image processing, thresholding is a well-known technique for image segmentation. Because of its wide applicability to other areas of the digital image processing, quite a number of thresholding methods have been proposed over the years. In this paper, we present a survey of thresholding techniques and update the earlier survey work by Weszka (Comput. Vision Graphics 62 Image Process 7, 1978, 259-265) and Fu and Mu (Pattern Recognit. 13, 1981, 3-16). We attempt to evaluate the performance of some automatic global thresholding methods using the criterion functions such as uniformity and shape measures. The evaluation is based on some real world images. 0 1988 Academic press, IX. f(K Y)-Let t E G be a threshold and B = {b,, b,} be a pair of binary gray levels and b,,b, E G. The result of thresholding an image function f(. , a) at gray level t is a

View PDFchevron_right
Binarisation of various images by detecting local thresholds with a validation test
Kacem Chehdi

[1991] IEEE Pacific Rim Conference on Communications, Computers and Signal Processing Conference Proceedings, 1991

The threshold method is one of the most important stages in the image pattern recognition chain when the object to be modeled does not contain texture. The authors present an efficient threshold method for a variety of images. It is based on the automatic detection of the different local thresholds of the image, taking into consideration the local properties of

View PDFchevron_right

Related topics

  • Information Systems
  • Applied Mathematics
  • Image Processing
  • Reinforcement Learning

    • Cited by

    Software Agent with Reinforcement Learning Approach for Medical Image Segmentation
    Chaw Seng Woo

    Journal of Computer Science and Technology, 2011

    Many image segmentation solutions are problem-based. Medical images have very similar grey level and texture among the interested objects. Therefore, medical image segmentation requires improvements although there have been researches done since the last few decades. We design a self-learning framework to extract several objects of interest simultaneously from Computed Tomography (CT) images. Our segmentation method has a learning phase that is based on reinforcement learning (RL) system. Each RL agent works on a particular sub-image of an input image to find a suitable value for each object in it. The RL system is define by state, action and reward. We defined some actions for each state in the sub-image. A reward function computes reward for each action of the RL agent. Finally, the valuable information, from discovering all states of the interest objects, will be stored in a Q-matrix and the final result can be applied in segmentation of similar images. The experimental results for cranial CT images demonstrated segmentation accuracy above 95%.

    View PDFchevron_right
    580 California St., Suite 400
    San Francisco, CA, 94104
    © 2025 Academia. All rights reserved