Texture Detection for Segmentation of Iris Images ASHEER KASAR BACHOO University of Kwa-Zulu Natal, South Africa and JULES-RAYMOND TAPAMO University of Kwa-Zulu Natal, South Africa

The idea of using the distinct spatial distribution of patterns in the human iris for person authentication is now a widely developing technology. Current systems rely on a set of basic assumptions in order to improve the accuracy and running time of the recognition process. The advent of a robust system implies a viable solution to a number of general problems. This paper focuses on a common yet difficult problem - the segmentation of eyelashes from iris texture. Tests give promising results when using grey level co-occurrence matrix (GLCM) approach. Categories and Subject Descriptors: I.4.6. [Image Processing and Computer Vision]: Segmentation - region growing and partition; I.5.3. [Pattern Recognition]: Clustering - algorithms, similarity measures General Terms: texture, classification, experimentation Additional Key Words and Phrases: iris, localization, GLCM, K-Means

1.

INTRODUCTION

The iris begins its formation in the 3rd month of gestation [Adler 1965]. By the 8th month, its distinctive pattern is complete. However, pigmentation and even pupil size increase as far up as adolescence [Kronfeld 1968]. The iris has a multilayered texture. This combination of layers and colour provide a highly distinctive pattern. An assortment of texture variations are possible. They include: —Contractile lines related to the state of the pupil —Crypts - irregular atrophy of the border layer. —Naevi - small elevations of the border layer. —Freckles - collections of chromatophores. —Colour variation - an increase in pigmentation yields darker coloured irides. Of the utmost importance in a biometric identification system is the stability and uniqueness of the object being analysed. Ophthalmologists [Flom and Safir 1987] and anatomists [Davson 1990], during the course of clinical observations, have noted that the irises of individuals are highly distinctive. This extends to the left and right eye of the same person. Repeated observations over a period of time have highlighted little variation in the patterns. Developmental biology has also provided evidence of the particular characteristics of the iris [Kronfeld 1968]. Although the general structure of the iris is genetically determined, the uniqueness of its minutiae is highly dependent on circumstances. As a result, replication is almost impossible. It has also been noted that, following adolescence, the iris remains stable and varies little for the remainder of the person’s life. Development is continuous during the early and adolescent years (pigmentation continues as well as an increase in pupil size) [Davson 1990; Kronfeld 1968]. Figure 1 shows an iris and the variety of its texture patterns. In 1936, Frank Burch, an ophthalmologist, proposed the idea of using iris patterns for personal identification. However, this was only documented by James Doggarts in 1949. The idea of iris identification for automated recognition was finally patented by Aran Safir and Leonard Flom in 1987. Although they had patented the idea, the two ophthalmologists were unsure as to a practical implementation for the system. They commissioned John Daugman to develop the fundamental algorithms in 1989. These algorithms were patented by Daugman in 1994 Asheer Kasar Bachoo, School of Computer Science, University of Kwa-Zulu Natal, Durban, 4041, e-mail - [email protected] Jules-Raymond Tapamo, School of Computer Science, University of Kwa-Zulu Natal, Durban, 4041, e-mail - [email protected] Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that the copies are not made or distributed for profit or commercial advantage, that the copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than SAICSIT or the ACM must be honoured. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or a fee. c 2005 SAICSIT ° Proceedings of SAICSIT 2005, Pages 111–118

112



Bachoo and Tapamo

Ciliary zone

¾

Collarette

¾

Pupillary zone

¾

Crypt

Pupil

Radial furrow

¾

Pupillary frill

-

Iris Sclera

Figure 1.

The iris

and now form the basis for all current commercial iris recognition systems. The Daugman algorithms are owned by Iridian Technologies and they are licensed to several other companies. 2.

RELATED WORK

The fundamental iris algorithm and system was developed by John Daugman. This was presented in his landmark paper [Daugman 1993] and subsequently updated [Daugman 2003]. The iris is segmented using integro-differential operators. To extract the rich details of the texture, he uses complex-valued 2D Gabor wavelets to extract discriminating information. Recognition is done by means of a test of statistical independence for two iris codes. A failure of the test implies a match. The matching system implements a normalized Hamming distance criteria. Features can be extracted using an application of Laplacian of Gaussian filters at different resolutions [Wildes 1997]. This proves to be efficient and consistent. The system uses edge maps and a Hough transform to locate the region of interest. A normalized correlation between the acquired iris representation and the stored one is employed for pattern matching. Feature extraction using zero-crossing representations of a 1D wavelet transform [Boles and Boashash 1998], the Haar wavelet [Ali and Hassanien 2003] and the Haar wavelet with a neural network for identification [K. Lim, K. Lee, O. Byeon and T. Kim 2001] are well documented. Multi-channel Gabor filtering [K. Lim, Y. Wang and T. Tan 2002], circular symmetric filters [L. Ma, Y. Wang and T. Tan 2002] and key local variations in 1D signals [L.Ma, T. Tan, Y. Wang and D. Zhang 2004] have also shown to be very promising for generating digital iris codes. There have also been a number of other interesting ideas for iris recognition. Multiresolution Independent Component Analysis (MICA) in [K. Bae, S Noh and J. Kim 2003] and application of the Hough transform for iris identification [W. Zorski, B. Foxon, J. Blackledge and M. Turner 2002] have produced some good results. A promising avenue has been an analysis of the Fourier spectra of the optical transmission binary models of human irises [A. Muron, P. Kois and J. Pospisil 2001]. Proceedings of SAICSIT 2005

Texture Detection for Segmentation of Iris Images using a New GLCM Feature



113

Very often, in real world problems, images do not demonstrate uniform intensities. Observation shows that these surfaces generally contain intensity variations that follow a particular pattern or periodicity. This is visual texture [C.H. Chen, L.F. Pau and P.S.P. Wang 1998]. A multitude of definitions are available for this characteristic of images - each one is particular to its domain and application. As a result, no formal definition exists. The human visual system can discriminate easily between different textures - it analyzes spatial, orientation, phase, frequency and colour information. Texture analysis in machine vision is time and space intensive. It is also sensitive to scale, resolution and orientation. The techniques available for machine learning consist of co-occurrence matrices, Markov random fields, wavelets and fractals, each one being adaptable to a particular situation [B. Jahne, H. Haubecker and P. Geibler 1999]. The fundamental iris recognition papers [Daugman 1993; Wildes 1997; Boles and Boashash 1998], while instrumental in providing rich knowledge for all subsequent endeavors, have not addressed the problem of interference caused by eyelashes. Iris texture is the region of interest (ROI) and it is required that we remove useless artifacts - such as eyelashes and eyelids - from the iris region. The notion of eyelashes as a texture can be observed in their displaying of a particular pattern and periodicity which is distinct from their surrounding areas. As a result, a texture segmentation approach is undertaken. An algorithm - void of texture classification - to detect eyelashes for accurate iris segmentation has been proposed in the literature [Kong and Zhang 2003]. It divides the problem into two possibilities - separable eyelashes and multiple eyelashes. Separable eyelashes are treated as edges. The image is convolved with a Gabor filter and thresholded to segment the eyelashes. The Gabor function is defined as G(x, u, σ) = exp{

x2 }cos(2πux) 2σ 2

(1)

where u is the frequency of the sinusoidal wave and σ is the standard deviation of the Gaussian. If the resultant value of a point falls below a threshold, it belongs to an eyelash. This can be stated as: f (x) ∗ G(v, u, σ) < K1

(2)

where K1 is a pre-determined threshold and ∗ is the convolution operator. The success of this process depends on a high intensity difference between iris pixels and eyelash pixels. Multiple eyelashes are modelled using an intensity variation model - eyelashes overlapping in a small area have a low intensity variation. If the variance of intensity in the area is below a threshold, the center of the window is labelled an eyelash pixel. This can be stated as: PN PN 2 i=−N j=−N (f (x + i, y + j) − M ) < K2 (3) 2 (2N + 1) where M is the mean intensity in the window and (2N +1)2 is the window size and K2 is a threshold. A connected components algorithm is also implemented to avoid misclassification of pixels. K1 and K2 are empirically determined parameters. The frequency distribution of iris images has been analyzed to determine occlusions by eyelids and eyelashes [L. Ma, T. Tan, Y. Wang and D. Zhang 2003]. While effective, the technique does not provide a solution to removing the eyelashes. 3.

IRIS LOCALIZATION

Segmentation of the iris from an image of the eye is sensitive to numerous factors - noise, uneven lighting and eyelid and eyelash interference. Although the pupil and iris borders can be modelled using two non-concentric circles, with the larger one forming a closed contour around the smaller one, detection of these two boundaries is not a simple problem. In a typical case, the outer iris boundary is a very soft gradient that standard edge functions do not detect. This boundary is sometimes only partially visible or, possibly, covered by eyelash. The details of the algorithm can be found in [Bachoo and Tapamo 2004]. 4.

EYELASH SEGMENTATION USING GREY LEVEL CO-OCCURRENCE MATRIX (GLCM) FEATURES

The technique presented by Kong and Zhang to segment eyelashes assumes a clear intensity distinction between iris and eyelash pixels. As a result, edge and variance operators are used for segmentation purposes. However, edge detectors are not always optimal and will fail when high texture variations are present or when the intensity gradient between edge and iris pixels is gradual rather then a step. There may also be cases of misclassification of pixels. The technique proposed in this paper uses texture analysis and pattern classification for segmentation. Proceedings of SAICSIT 2005

114



Bachoo and Tapamo

Figure 2.

An iris image

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

Figure 3. Segmentation results using different features computed from GLCM - (a)-(d) K-means clustering; (e)-(h) Fuzzy C-Means clustering. The features used for classification are (for each image from left to right) energy, entropy, correlation and mean. Co-occurrence features are an effective texture descriptor [R.M. Haralick, K. Shanmugam and I. Dinstein 1973]. It estimates the image properties to second-order statistical features. A GLCM (Grey Level Co-occurrence Matrix) element pθ,d (i, j) is the joint probability of the grey level pairs i and j in a given direction θ separated by distance of d pixels. G denotes the number of grey levels. For each ROI (Region Of Interest) in this work, the feature computed using symmetrical GLCMs is a new one defined as follows: F=

G X

i pθ,d (i, j)( ) j i,j=1

(4)

The texture features computed from the GLCM are not all useful. A number of test runs have shown that the most reliable feature for segmenting regions in an iris image is the one defined above. A possible reason for this is the small size of the ROI - it is a disc with an inner radius of approximately 40 pixels and an outer radius of about 110 pixels. Features such as entropy, energy and correlation are difficult to capture effectively within a small region with a large number of grey levels. Selection of features for texture discrimination is application dependent. A feature set for describing grass will not classify cotton threads although it may produce useful results [Q. Zhang, J. Wang, P. Gong and P. Shi 2003]. Texture discrimination for eyelash segmentation is, thus, dependent on a careful analysis of the problem and preliminary implementation and test results. Using a GLCM to classify textures is sensitive to many factors. Selection of a window size is important for capturing the relevant statistics - a small window will omit important grey levels and structure; a large window size will include different texture areas and, thus, add incorrect statistics to the GLCM. The number of grey levels in the image texture also affect the feature extraction process - a large number of grey levels will imply exhaustive computational time and also provide poor relationships between pixels of the same texture due to the high randomness present if there are too few pixels to examine. A reduced set of grey levels will improve Proceedings of SAICSIT 2005

Texture Detection for Segmentation of Iris Images using a New GLCM Feature



115

the computational burden but destroy certain textural characteristics [R.M. Haralick, K. Shanmugam and I. Dinstein 1973]. These considerations must be taken in order to design an accurate and robust image processing system. While an automatic system is highly desirable, it is difficult to know all the variables in the problem. A good way to approach the problem is to do a manual analysis of the images and problem and a few test runs to ascertain reliable parameters. These can then be built into the system so that most situations can be dealt with. Figure 2 shows a test image of an iris followed by Figure 3 with some segmentation results using different GLCM features. To produce meaningful results, different parameter values for d and θ must be used. This improves the statistical details of the texture and helps generalize a texture definition for a particular class, improving the class uniqueness. However, unnecessary parameters can also influence and degrade the feature class. 5.

UNSUPERVISED FEATURE CLASSIFICATION

The segmentation of eyelashes is a problem of unsupervised classification. In essence, we are required to partition the set of feature vectors into classes that differ from each other as much as possible. Concurrently, the vectors in the same class must differ as little as possible. The number of distinct texture classes in the region of interest is unknown. Ideally, we require a two class classifier that separates iris texture and any other textures. The K-means [MacQueen 1967] and fuzzy C-means [Dunn 1974; Bezdek 1981] algorithms are implemented. There are a multitude of textures possible when an image of the eye is captured - skin, eyelash, sclera, iris and pupil. The texture variations are random due to the following factors: (1) (2) (3) (4) 5.1

Uneven illumination caused by the image acquisition process. The uniqueness of every iris pattern. The presence of multiple, single or no eyelashes. The pupil dilating or constricting, which changes the density of the iris pattern. K-means Clustering

The MacQueens k-means algorithm is a popular clustering algorithm in which the number of clusters (k) is known. It is an iterative process that assigns patterns to the closest cluster using a distance function (such as the Euclidean distance measure). The basic algorithm is described below. (1) (2) (3) (4) (5) 5.2

Define the number of clusters k. Initialize (randomly) the clusters prototypes pi (i = 1..k). For each pattern x, assign x to the nearest cluster pi (i = 1..k). Recompute pi (i = 1..k). Repeat steps 3 and 4 until the prototypes do not change. Fuzzy C-means Clustering

Fuzzy clustering allows data to belong to more than one class. This is reflected by their degree of membership in a particular cluster. It is based on the minimization of the objective function Jm =

C N X X

2 ,1≤m≤∞ um ij k xi − vj k

(5)

i=1 j=1

where m is a real number greater than 1 (called the fuzzification factor). uij is the degree of membership of xi in the cluster j where 0 ≤ uij ≤ 1. xi is the i th component in a d-dimensional data set (vectors). vj is the d-dimension center of cluster j and k ∗ k is the Euclidean norm. C denotes the number of clusters and N the number of pattern vectors. Fuzzy partitioning is an iterative optimization process. The degree of membership (uij ) and the cluster centers vj are computed by the following equation: 1

uij = PC

2 kxi −vj k m−1 k=1 ( kxi −vk k )

vj =

PN

i=1

PN

um ij ¦ xi

i=1

um ij

(6)

(7)

The stopping criterion is | ukij − uk−1 |< ², where ² is a threshold between zero and one. The algorithm is as ij follows: Proceedings of SAICSIT 2005

116



Bachoo and Tapamo

Figure 4.

Figure 5.

Figure 6.

3 test images

Test results using K-Means clustering

Test results using fuzzy C-means clustering

(1) (2) (3) (4) (5) (6)

Select the values of ² and c and initialize the counter k to 0 (k ← 0). Obtain (randomly) the initial membership matrix, U (0) . k ←k+1 Compute the centroids vj (j = 1, ..., c) using equation (7). Compute the membership matrix U (k) using equation (6). If || U (k) − U (k−1) ||< ² then stop else go to step 3.

6.

EXPERIMENTAL RESULTS AND DISCUSSION

The number of texture classes in an iris image can vary from 2 to 5. Thus, ideally, the maximum number of classes is 5. A reasonable parameter for the clustering algorithms is 3. Input images are of size 320 × 280 pixels with 256 grey levels. 12 symmetrical GLCMs are generated for θ ∈ {0, 45, 90, 135} and d ∈ {1, 3, 5} with a reduced set of 64 grey levels. A window size of 21 × 21 pixels (non-overlapping sub-images) is implemented. Smaller windows can be implemented for finer segmentation but classification results may be poorer. Test images are used from the CASIA iris database [CASIA Iris Image Database ]. 30 images were tested using the new GLCM feature and the parameters described above. The K-Means and fuzzy C-Means were both implemented on each image for comparison. To test the accuracy of the proposed system, these outputs were manually inspected and poor or unsatisfactory results were discarded. The K-Means algorithm effectively segmented 22 out of 30 images while the fuzzy C-Means clustering approach performed slightly better with 24 Proceedings of SAICSIT 2005

Texture Detection for Segmentation of Iris Images using a New GLCM Feature

Number of images Segmented Rejected Accuracy Table I.

K-Means 30 22 8 73.3%



117

Fuzzy C-Means 30 24 6 80%

Clustering results

images being well segmented. Figure 4 shows 3 test images, with their segmentation outputs in Figure 5 and Figure 6. Table I shows the clustering results. The sample space for the above experiments comprised of images with multiple eyelashes, single eyelashes, uneven illumination and various texture details. The algorithm fails in the following instances: —When the texture of the eyelash zone is indiscernible from that of another zone. —In the presence of uneven lighting. —Poor parameter selection eg. the texture window size or the number of clusters for the clustering. Criteria for manual rejection or acceptance of clustering results were based on over segmentation and correct classification. Clustering outputs produced an image with colours for different zones. These zones were compared to the original image to check for misclassification. Misclassified pixels were similar objects with different colours or different objects with the same colour. 7.

FUTURE DEVELOPMENTS

The new GLCM texture measure for segmenting iris images is a promising technique. As mentioned earlier, a number of factors affect the performance of the segmentation algorithm. It is hoped that a fully automatic approach can be designed for establishing parameters for window size, number of grey levels and distinguishing features for a particular texture class in any application. Our current experiments also entail Gabor filters [P. Kruizinga, N. Petkov and S.E. Grigorescu 1999], discrete wavelet transform (DWT) [Mallat 1989] and Markov Random Fields (MRF) [Chellappa and Chatterjee 1985] for texture discrimination. A combination of some of these methods will be performed for optimal segmentation. Currently, the system is fully unsupervised. A supervised approach is being implemented by performing training on iris images in order to compute a feature library to assist in pattern recognition. This will also provide statistics for cluster validation and finer segmentation. It is also hope that image enhancement can be improved in order to address uneven lighting and indiscernible texture boundaries. 8.

CONCLUSION

A texture analysis approach to improve iris segmentation for person identification has been presented. A new feature derived from the GLCM has been proposed as effective for texture classification. The test results show that the proposed approach is very promising and can help improve the recognition rate of iris systems by removing noisy artifacts. ACKNOWLEDGMENTS

Portions of the research in this paper use the CASIA iris image database collected by the Institute of Automation, Chinese Academy of Sciences. The financial assistance of the National Research Foundation (NRF) towards this research is hereby acknowledged. Opinions expressed and conclusions arrived at are those of the authors and not necessarily to be attributed to the NRF. Armscor (SA) is also acknowledged for their financial support towards this project. REFERENCES A. Muron, P. Kois and J. Pospisil. 2001. Identification of persons by means of the fourier spectra of the optical transmission binary models of human irises. Optics Communications 192, 161–167. Adler, F. 1965. Physiology of the eye. MO:Mosby, St. Louis. Ali, J. and Hassanien, A. 2003. An iris recognition system to enhance e-security environment based on wavelet theory. Advanced Modelling and Optimization 5, 2. B. Jahne, H. Haubecker and P. Geibler. 1999. Handbook of Computer Vision and Applications - Volume 2. Academic Press. Bachoo, A. and Tapamo, J.-R. 2004. A segmentation method to improve iris-based person identification. In Africon 2004. Vol. 1. 403–408. Proceedings of SAICSIT 2005

118



Bachoo and Tapamo

Bezdek, J. 1981. Pattern Recognition with Fuzzy Object Function. Plenum Press. Boles, L. and Boashash, B. 1998. A human identification technique using images of the iris and wavelet transform. IEEE Trans. on Signal Processing 46, 4, 1185–1188. CASIA Iris Image Database. http://www.sinobiometrics.com. C.H. Chen, L.F. Pau and P.S.P. Wang. 1998. The Handbook of Pattern Recognition and Computer Vision (2nd Edition). World Scientific Publishing Co. Chellappa, R. and Chatterjee, S. 1985. Classification of textures using gaussian markov random fields. IEEE Trans. Acous., Speech Signal Proc. 33, 4, 959–963. Daugman, J. 1993. High confidence visual recognition of persons by a test of statistical independance. IEEE Trans. Pattern Analysis Mach. Intell. 15, 11, 1148–1961. Daugman, J. 2003. The importance of being random: stastical principles of iris recognition. Pattern Recognition 36, 2, 279–291. Davson, H. 1990. Davson’s Physiology of the Eye. MacMillan, London. Dunn, J. 1974. A fuzzy relative of the isodata process and its use in detecting compact well separated clusters. Journal of Cybernetics 3, 32–57. Flom, L. and Safir, A. 1987. Iris recognition system. K. Bae, S Noh and J. Kim. 2003. Iris feature extraction using independent component analysis. In AVBPA. 838–844. K. Lim, K. Lee, O. Byeon and T. Kim. 2001. Efficient iris recognition through improvement of feature vector and classifier. ETRI Journal 23, 2, 61–70. K. Lim, Y. Wang and T. Tan. 2002. Iris recognition based on multichannel gabor filtering. In Fifth Asian Conference on Computer Vision. Vol. 1. 279–283. Kong, W. and Zhang, D. 2003. Detecting eyelash and reflection for accurate iris segmentation. International Journal of Pattern Recognition and Artificial Intelligence 17, 6, 1025–1034. Kronfeld, P. 1968. The gross embryology of the eye. The Eye 1, 1–66. L. Ma, T. Tan, Y. Wang and D. Zhang. 2003. Personal identification based on iris texture analysis. IEEE Trans. Pattern Analysis Mach. Intell. 25, 12, 1519–1533. L. Ma, Y. Wang and T. Tan. 2002. Iris recognition using circular symmetric filters. In ICPR. Vol. 2. 414–417. L.Ma, T. Tan, Y. Wang and D. Zhang. 2004. Efficient iris recognition by characterizing key local variations. IEEE Transactions on Image Processing 13, 6, 739–750. MacQueen, J. 1967. Some methods for classification and analysis of multivariate observations. Proceedings of the 5th Berkeley Symposium-1 , 281–297. Mallat, S. 1989. A theory of multiresolution signal decomposition. IEEE Trans. on Pattern Analysis and Machine Inteligence 11, 7, 674–693. P. Kruizinga, N. Petkov and S.E. Grigorescu. 1999. Comparison of texture features based on gabor filters. In Proceedings of the 10th International Conference on Image Analysis and Processing. 142–147. Q. Zhang, J. Wang, P. Gong and P. Shi. 2003. Study of urban spatial patterns from spot panchromatic imagery using textural analysis. Int. J. Remote Sensing 24, 21, 4137–4160. R.M. Haralick, K. Shanmugam and I. Dinstein. 1973. Texture features for image classification. IEEE Trans. System Man. Cybernat. 8, 6, 610–621. W. Zorski, B. Foxon, J. Blackledge and M. Turner. 2002. Fingerprint and iris identification method based on the hough transform. In Proceedings of IMA Third Conference on imaging and Digital Image Processing. Wildes, R. 1997. Iris recognition: an emerging biometric technology. Proceedings of the IEEE 85, 9, 1348–1362.

Proceedings of SAICSIT 2005

Texture Detection for Segmentation of Iris Images - CiteSeerX

Asheer Kasar Bachoo, School of Computer Science, University of Kwa-Zulu Natal, ..... than 1 (called the fuzzification factor). uij is the degree of membership of xi.

332KB Sizes 5 Downloads 314 Views

Recommend Documents

CLUSTERING of TEXTURE FEATURES for CONTENT ... - CiteSeerX
storage devices, scanning, networking, image compression, and desktop ... The typical application areas of such systems are medical image databases, photo ...

Texture Measures for Improved Watershed Segmentation of Froth ...
ore type changes) to each float cell. Typically, changes to the input variables are made by an experienced operator, based on a visual inspection of the froth. These operators look at features such as: bubble size, froth velocity, froth colour and fr

Geometry Motivated Variational Segmentation for Color Images
In Section 2 we give a review of variational segmentation and color edge detection. .... It turns out (see [4]) that this functional has an integral representation.

Affine layer segmentation and adjacency graphs for vortex ... - USC IRIS
modeling the motion as affine. The underlying physical properties of meteorological phenomena like cloudy structures are complex. Our system does not take into account any underlying model of the meteorological phenomena, instead we use the informati

An Effective Segmentation Method for Iris Recognition System
Biometric identification is an emerging technology which gains more attention in recent years. ... characteristics, iris has distinct phase information which spans about 249 degrees of freedom [6,7]. This advantage let iris recognition be the most ..

Segmentation of Connected Chinese Characters Based ... - CiteSeerX
State Key Lab of Intelligent Tech. & Sys., CST ... Function to decide which one is the best among all ... construct the best segmentation path, genetic algorithm.

Texture based segmentation of remotely sensed ...
The exact purpose of image segmentation is either to extract the outlines of different regions in the ...... Simplicity of these metrics may in some cases cause mis-.

Semantic-Shift for Unsupervised Object Detection - CiteSeerX
notated images for constructing a supervised image under- standing system. .... the same way as in learning, but now keeping the factors. P(wj|zk) ... sponds to the foreground object as zFG and call it the fore- ..... In European Conference on.

Profile Injection Attack Detection for Securing ... - CiteSeerX
to Bamshad Mobasher for inspiring and encouraging me to pursue an academic career in computer science. His thoroughness and promptness in reviewing my ...

Profile Injection Attack Detection for Securing ... - CiteSeerX
6.4.2 Robustness Comparison Against Nuke Attacks . . . . . . . . . . . . . . 34. 7 Defense ..... As a result, we have focused on profile analytics data and attribute ...... Data Mining: Practical machine learning tools and techniques, 2nd. Edition. M

Pollination of Oncocyclus irises (Iris: Iridaceae) by Night ... - CiteSeerX
May 25, 2005 - haynei on Mt. Gilboa was very high (> 90%) throughout all floral life stages. .... compatibility, (3) no pollination of covered flowers as a control.

Pattern recognition techniques for automatic detection of ... - CiteSeerX
Computer-aided diagnosis;. Machine learning. Summary We have employed two pattern recognition methods used commonly for face recognition in order to analyse digital mammograms. ..... should have values near 1.0 on the main diagonal,. i.e., for true .

Abdominal Multi-Organ Segmentation of CT Images ... - Springer Link
Graduate School of Information Science and Engineering, Ritsumeikan University,. 1-1-1, Nojihigashi, .... the segmented region boundaries Bs of the “stable” organs, we estimate b. ∗ by b. ∗. = argmin .... International Journal of Computer As-

A Segmentation Method to Improve Iris-based Person ...
unique digital signature for an individual. As a result, the stability and integrity of a system depends on effective segmentation of the iris to generate the iris-code.

Joint Segmentation and Spiculation Detection for Ill ...
a Department of Electrical and Computer Engineering, Virginia Tech, Arlington, ... The pixel-level mass and non-mass class labeling and segmentation works as.

Segmentation of Mosaic Images based on Deformable ...
in this important application domain, from a number of points of view including ... To the best of our knowledge, there is only one mosaic-oriented segmentation.

segmentation techniques of microarray images – a ...
Cochin University of Science and Technology. By. JISHNU L. REUBEN ... analysis of gene expression. A DNA microarray is a multiplex technology used in.

Protein Word Detection using Text Segmentation Techniques
Aug 4, 2017 - They call the short consequent sequences (SCS) present in ..... In Proceedings of the Joint Conference of the 47th ... ACM SIGMOBILE Mobile.

Review of Iris Recognition System Iris Recognition System Iris ... - IJRIT
Abstract. Iris recognition is an important biometric method for human identification with high accuracy. It is the most reliable and accurate biometric identification system available today. This paper gives an overview of the research on iris recogn

NetPiler: Detection of Ineffective Router Configurations - CiteSeerX
A simple example configuration of inter-domain routing. A line between two .... its settlement-free peers, one can apply a route filter on BGP sessions with each of ...

Pattern recognition techniques for automatic detection of ... - CiteSeerX
Moreover, it is com- mon for a cluster of micro–calcifications to reveal morphological features that cannot be clearly clas- sified as being benign or malignant. Other important breast .... mammography will be created because of the data deluge to

Joint Segmentation and Spiculation Detection for Ill ...
[2] extended the probabilistic model based method by further analyzing the steepest change of the cost functions. Their method was found to be able to further include some of ill-defined mass boundaries. Te Brake and Karssemeijer [3] proposed a discr