Vol. 19 (2010):
Abstracts of Papers
A Color- and Texture-based Image Segmentation Algorithm.
MGV vol. 19, no. 1, 2010, pp. 3-18.
Image segmentation is a classic inverse problem which consists
in obtaining a compact, region-based description of the image scene
by decomposing it into meaningful or spatially coherent regions
sharing similar attributes. Because a color image can provide more
perceptual information, color image segmentation is being paid more
and more attention. In this paper, we propose a new approach to
color image segmentation that is based on low-level features of
color and texture. The approach is aimed at segmentation of natural
scenes where the color and texture of each segment do not
typically exhibit uniform statistical characteristics. Firstly,
local color composition is described in terms of spatially adaptive dominant
colors by using the Gibbs random field, and the color image
is segmented into regions according to the local color composition.
Secondly, the texture characteristics of the grayscale component are
described by utilizing the Steerable filter, and the grayscale
component of color image is cut into flat regions and non-flat
regions. Thirdly, the local color composition and texture
characteristics are combined to obtain an overall crude
segmentation. Finally, an elaborate border refinement procedure is
used to obtain accurate and precise border localization by
appropriately combining color-texture features with the Normalized
Cuts. The experimental results demonstrate that the color image
segmentation results of the proposed approach exhibit favorable
consistency in terms of human perception.
Image segmentation, Gibbs random
field, Steerable filter, Normalized Cuts.
Youssef B. A.:
Image Segmentation Using Streamlines Analogy.
MGV vol. 19, no. 1, 2010, pp. 19-31.
This paper presents a novel method for digital image segmentation based on the analogy between
streamlines in fluid dynamics
and isophote lines in digital images. The segmentation problem is reformulated so that the image
intensity corresponds to the stream function for a two-dimensional, incompressible flow,
and image intensity gradients are represented as the fluid velocity vector.
Segmentation is effected by computing the streamlines by solving a coupled system
of ordinary differential equations using a fourth-order Runge-Kutta method.
Selection of the initial starting point for segmentation is based on color homogeneity in terms of
local color gradient, and on variance. The effectiveness of the developed segmentation method is
demonstrated through a number of case studies, ranging from gray level to colored images.
Image segmentation, streamlines, stream function.
Roy K., Bhattacharya P.:
Iris Recognition Using Genetic Algorithms and Asymmetrical SVMs.
MGV vol. 19, no. 1, 2010, pp. 33-62.
With the increasing demand for enhanced security, iris biometrics-based personal
identification has become an interesting research topic in the field of pattern recognition.
While most state-of-the-art iris recognition algorithms are focused on preprocessing iris images,
important new directions have been identified recently in iris biometrics research.
These include optimal feature selection and iris pattern classification. In this paper,
we propose an iris recognition scheme based on Genetic Algorithms (GAs) and asymmetrical Support Vector
Machines (SVMs). Instead of using the whole iris region, we elicit the iris information between the
collarette and the pupillary boundaries to suppress effects of eyelids and eyelashes occlusions, and pupil dilation,
and to minimize the matching error. To select the optimal feature subset together with increasing the overall
recognition accuracy, we apply GAs with a new fitness function. The traditional SVMs are modified into
asymmetrical SVMs to handle: (1) highly unbalanced sample proportion between two classes,
and 2) different types of misclassification error that lead to different misclassification losses.
Furthermore, the parameters of SVMs are optimized in order to improve the generalization performance.
The proposed technique is computationally effective, with recognition rates of 97.80% and 95.70% on the
Iris Challenge Evaluation (ICE)
and the West Virginia University (WVU) iris datasets, respectively.
Biometrics, iris recognition, asymmetrical support vector machines,
collarette area localization, genetic algorithms.
Surfaces Filling Polygonal Holes with G1 Quasi G2 Continuity.
MGV vol. 19, no. 1, 2010, pp. 63-96.
Two constructions of surfaces filling polygonal holes in piecewise B-spline bicubic surfaces with tangent
plane continuity are described. The filling surfaces are obtained by minimization of functionals which impose penalty on curvature discontinuities. One of the functionals
is a quadratic form, while the other functional is defined with
a parameterization-independent formula. The resulting surfaces may be used
instead of class G2 surfaces in practical applications; the
penalty approach enables simplification of the construction, and
reduction in the degree of patches filling the hole from (9,9) to (5,5)
without any visible quality degradation.
The notion of class Gn quasi Gm surfaces, i.e. class Gn surfaces
optimized to approximate class Gm surfaces, is proposed.
Filling polygonal holes, tangent plane continuity,
curvature continuity, compatibility conditions, surface shape optimization,
Liu S., Li J.:
Preserving Zeros in Surface Construction using Marching Cubes.
MGV vol. 19, no. 1, 2010, pp. 97-123.
In surface construction, existing marching cubes (MC) methods require
sample values at cell vertices to be non-zero after
thresholding, or modify them otherwise. The modification may
introduce problems in the constructed surface, such as topological changes, representation errors, and preference for positive or negative values.
This paper presents a generalized MC algorithm.
It constructs surface patches by exploiting cycles in cells without changing the sample values at vertices, and thus allows cell vertices with zero sample values to lie on the constructed surface.
The simulation results show that
the proposed Zero-Crossing MC method preserves better topologies of implicit surfaces
that pass through cell vertices, and represents the surfaces more accurately.
Its efficiency is comparable to existing MC methods in constructing surfaces.
Surface construction, Marching cubes, Zero-crossing.
Fast Object Detection Using Steiner Tree.
MGV vol. 19, no. 2, 2010, pp. 127-142.
We propose an approach to speed-up object detection, with an emphasis on settings where multiple
object classes are detected. Our method uses a segmentation algorithm to select a small number of image regions
on which to run a classifier. Compared to the classical sliding window approach, a significantly smaller number
of rectangles is examined, which yields significantly faster object detection.
Further, in the multiple object class setting, we show that the computational cost of segmentations
can be amortized across objects classes, resulting in an additional speedup. At the heart of our approach
is reduction to a directed Steiner tree optimization problem, which we solve approximately in order to
select the segmentation algorithm parameters. The solution gives a small set of segmentation strategies
that can be shared across object classes. Compared to the sliding window approach,
our method results in two orders of magnitude fewer regions considered, and significant (10-15x)
computational time speedups on challenging object detection datasets
(LabelMe and StreetScenes) while maintaining comparable detection accuracy.
Object, Detection, Recognition, Steiner tree.
Own H. S.:
Improvement in Image Denoising Technique Based on Dual -Tree Wavelet Transform and Multiresolution Local Contrast Entropy.
MGV vol. 19, no. 2, 2010, pp. 143-157.
The paper proposes an improvement in
image denoising using Dual-Tree Complex Wavelet Transform (DT-CWT).
Depending on the probability distribution of the noise in the
wavelet coefficients, an adaptive threshold estimation algorithm is
introduced. The threshold enables the proposed algorithm to
adapt to unknown smoothness of the noisy images. The algorithm
looks at the local contrast entropy of a complex wavelet coefficient
instead of its magnitude in order to remove the noise from the image.
Simulation results show improved performance of our image
denoising method compared to other popular denoising
algorithms, such as VisuShrink, Wiener2, ProbShrink, and our previous
work based on DWT.
Image Denoising, Dual Tree Complex Wavelet, Multiresolution Local
Gedda M., Öfverstedt L.-G., Skoglund U., Svensson S.:
Image Processing System for Localising Macromolecules in Cryo-Electron Tomography.
MGV vol. 19, no. 2, 2010, pp. 159-184.
A major challenge in today's molecular biology research is to understand the interaction between proteins at the molecular level. Cryo-electron tomography (ET) has come to play an important role in facilitating objective qualitative experiments on protein structures and their interaction. Various protein conformation structures can be qualitatively analysed as complete galleries of proteins are captured by ET. To facilitate fast and objective macromolecular structure analysis procedures, image processing has become a crucial tool. This paper presents an image processing system for localising individual proteins from in vitro samples imaged by ET. We have evaluated the system using simulated data as well as experimental data.
Fuzzy set, watershed segmentation, distance transform, proteins.
Smietanski J., Tadeusiewicz R.:
Discriminatory Power of Co-Occurrence Features in Perfusion CT Prostate Images.
MGV vol. 19, no. 2, 2010, pp. 185-199.
This paper presents an algorithm used to improve the
effectiveness of early prostate cancer (PCa)detection. The
necessity for using such a computational method lies in the fact
that although perfusion computed tomography (p-CT) is considered a
good technique for the detection of early PCa, the p-CT prostate
images are very difficult to interpret manually by radiologists.
We hereby propose a methodology for computational analysis of p-CT
prostate images based on textural coefficients derived from
co-occurrence matrices and their 21 coefficients. The selection of
only a few of the considered features ensures the necessary balance
between matching set of already known images and new, not yet clear
The proposed algorithm for automatic differentiation of the healthy
area of the image from the cancerous region was tested on a set of
59 prostate images. Although the results were not entirely
satisfactory (86% correct recognitions), this method may be
considered as the base for the development of a better algorithm.
prostate cancer, perfusion computed tomography, medical image analysis, pattern recognition.
Wood B.A., Lee J.K.,
Maskey M., Newman T.S.:
Higher Order Approximating Normals and Their Impact on Isosurface Shading
MGV vol. 19, no. 2, 2010, pp. 201-221.
Two alternatives to the standard (central differencing) method for estimating
normals of Marching Cubes isosurfaces are considered. The methods are based on
higher order approximations of dataset gradients. Of primary concern here are
the effects of these methods on rendering quality, which is evaluated here
through pixel-by-pixel comparisons of typical-fidelity isosurfaces versus
high-fidelity rendering achievable from analytically derived formulae. The
evaluations also consider effects of noise on isosurface rendering quality for
renderings based on standard versus higher order gradients.
Volume Visualization, Isosurfaces,
Marching Cubes, Shading & Illumination.
Kit D.L.H., Suandi
Reconstruction of 3D Surface From 2D Images Using Five Lighting Sources.
MGV vol. 19, no. 2, 2010, pp. 223-242.
This paper proposes a novel method for reconstructing and acquiring 3D geometry information on an inspected surface based on 2D images. In this research, multiple light sources and a single camera setup are used to capture several 2D images in order to reconstruct the a 3D surface model. The approach consists of two major stages to obtain the 3D geometry information: (1) the shade of the images will be used to get the surface gradient which will finally be used to construct the surface gradient map, and (2) the shadow of the object will be approximated in order to reconstruct the step height (edge) of the object.
Surface gradient mapping, shadow correction, five light sources illumination setting.
Special Issue on Image Databases
Special Issue Editor: J.L. Kulikowski.
Nikos Papadakis, Nikos Doulamis and Anastasios Doulamis:
Hierarchical Graph-Based Media Content Representation for Real Time
Search in Large Scale Multmedia Databases.
MGV vol. 19, no. 3, 2010, pp. 247-263.
This paper presents a system architecture and the appropriate algorithms
for confidential searching of digital multimedia libraries. The proposed scheme uses
the Middleware service layer that allows pre-processing of raw content with
the technology owned by the Search Engine, without compromising the security of the original architecture in any way. The specific search algorithms described are a hierarchical graph structure algorithm for preprocessing, and a backtracking search algorithm
that achieves good real-time performance (speed, and precision-recall values) under
the given security constraints.
Multimedia Databases, image
process, computer vision.
A. Lisowska, T. Kaczmarzyk:
JCURVE --- Multiscale Curve Coding via Second Order Beamlets.
MGV vol. 19, no. 3, 2010, pp. 265-281.
The paper presents an algorithm JCURVE for compression of binary
images with linear or curvilinear features, which is a kind of
generalization of the JBEAM coder. The proposed algorithm is based
on second order beamlet representation, where second order beamlets
are defined as hierarchically organized segments of conic curves.
The algorithm can compress images in both a lossy and losless way,
and it is also progressive. The experiments performed on benchmark
images have shown that the proposed algorithm significantly
outperforms the known JBIG2 standard and the base JBEAM algorithm
both in losless and lossy compression. It is characterized,
additionally, by the same time complexity as JBEAM, namely O(N2log2 N) for image of size N × N pixels.
Image compression, second order beamlets, curve coding.
W.W. Koczkodaj, A. Przelaskowski and K.T. Szopinski:
Medical Knowledge Mining from Image Data -- Synthesis of Medical Image Assessments for Early Stroke Detection.
MGV vol. 19, no. 2, 2010, pp. 283-.
The key issue of this study is synthesis
of medical images and expert knowledge
for early detection of a medical condition, such as stroke or cancer.
Such synthesis is a missing link for making decisions
during the diagnostic process.
Knowledge mining in image databases can be enhanced by computing
the relative importance of image features using pairwise comparisons.
Computed weights can be systematically used for synthesis of various
image features present in the same or different images.
knowledge mining, image data, pairwise comparisons,
inconsistency analysis, early stroke detection.
Visual Retrieval of Documents Based on Their
Multi-Aspect Utility Assessment.
MGV vol. 19, no. 3, 2010, pp. 299-320.
Differences between textual and visual documents retrieval
problems are described. It is shown that retrieval of visual
documents in experimental data bases requires assessment of image
utility and taking it into consideration. A definition of
multi-aspect image usefulness measure is proposed. A multi-aspect
measure of similarity of images based on their quantitative and/or
qualitative features is also proposed. The general concept is
illustrated by examples of using morphological spectra as a source
of parameters useful in the assessing similarity of some classes of
biomedical images. The basic structure and properties of an
Image Analysis and Selection System (IASS) are presented as
an example of practical realization of the visual documents
visual documents retrieval, image utility measure, image similarity measure, morphological spectra, analysis of
Dulewicz A., Jaszczak P., Kupis P.:
Application of a Pathomorphological Image Database in
Computer-Aided Cytological Examinations.
MGV vol. 19, no. 3, 2010, pp. 321-337.
The paper deals with the problem of early detection of bladder
cancer based on non-invasive, voided urine cytological
investigations. In spite of the diagnostic potential of the method
for discovering malignancy associated changes in cells before they
start to form a tumour, cytological tests seem to be underestimated
by physicians, as there is a common view that their sensitivity,
especially in early stages of the cancer, is relatively low. We
depict here just one, but significant, direction of our work that
aims to support the cytopathologist in making the diagnosis more
accurate and reliable. The key idea relies on classification of
adaptive smear objects by searching for similar patterns in a
flexible pathomorphological image database using content-based image
retrieval technology (CBIR).
bladder cancer, digital cytology, image processing, content-based image
Orientation and Pose estimation of Panoramic Imagery.
MGV vol. 19, no. 3, 2010, pp. 339-363.
In a database of geo-referenced images, determining the exact
position of each panorama is an important step in order to ensure
the consistency of visual information. This paper addresses the
problem of camera pose recovery from spherical (360o) panoramas.
The 3D information is extracted from a database of panoramic
images sparsely distributed over a scene of interest. We present a
two-stage algorithm to recover the position of omni-directional
cameras using pair wise essential matrices. First, all rotations
with respect to the world frame are found using an incremental
bundle adjustment procedure, thus achieving what we call
panorama alignment. Full camera positions are then computed
using bundle adjustment. During this step, the previously computed
panorama orientations, used to feed the global optimization process,
can be further refined. Results are shown for indoor and outdoor
omni-directional panorama; pose estimation; bundle adjustment; structure from
- Karasulu B., Balli
Image Segmentation Using Fuzzy Logic, Neural Networks and Genetic Algorithms: Survey and Trends.
MGV vol. 19, no. 4, 2010, pp. 367-409.
- Image segmentation is a fundamental process employed in many
applications of pattern recognition, video analysis, computer vision
and image understanding in order to allow further image content
exploitation in an efficient way. It is often used to partition an
image into separate regions. As recent trends in image segmentation
show, the use of artificial and/or computational intelligence (AI
and/or CI) techniques has become more popular as an alternative to
the conventional techniques. In this paper, we present an extensive
and comprehensive review of the image processing area for advanced
researchers. This study introduces the theoretical fundamentals of
image segmentation using AI and/or CI techniques based on fuzzy
logic (FL), genetic algorithm (GA) and artificial neural networks
(ANN). Besides, this survey examines the applications of these
techniques in different image segmentation areas. In the literature,
these techniques are used as an interpretation tool for
segmentation. In our study, these tools are focused on because of
their capabilities, such as robustness, segmentation accuracy and
low computational costs. Moreover, we review 56 remarkable studies
from the last decade (i.e., the years between 2001 and 2010), which
involve different image segmentation approaches using FL, GAs, ANNs
and hybrid intelligent systems (HISs). In our state-of-the-art
survey, the comparison of the reviewed papers in related categories
is made based on both the corresponding properties of segmentation
as well as performance evaluation of the related method proposed in
a given reviewed paper. The results and recent trends are also
Key words: image segmentation, neural networks, fuzzy logic, genetic algorithms, clustering.
- Pawar V. N., Talbar S. N.:
Hybrid Machine Learning Approach for Object Recognition: Fusion of Features and Decisions.
MGV vol. 19, no. 4, 2010, pp. 411-428.
- Object recognition is
considered to be a predominant basic issue in computer vision. It is
a challenging issue against inconsistent illumination, partial
occlusion, changing background and shifting viewpoint, because
considerable variations are exhibited by diversified real world
patterns. The virtue of feature fusion lies in its reliability and
capability for object recognition in terms of actual redundancy and
complementary information. In this paper, we have developed an
efficient hybrid approach using scale invariant features and machine
learning techniques for object recognition. We extract the scale
invariant features, namely color, shape and texture of the objects,
separately with the aid of suitable feature extraction techniques.
Then, we integrate the color, shape and texture features of the
objects at the feature level, so as to improve the recognition
performance. The fused feature set serves as a pattern for the
forthcoming processes involved in the developed approach.
Subsequently, we hybridize the process of object recognition by
combining the pattern recognition algorithms like Support Vector
Machine, Discriminant Canonical Correlation, and Locality Preserving
Projections. Obviously, with three different pattern recognition
algorithms employed, we are likely to get three distinct or
identical results enumbered with false positives. So in order to
reduce the number of false positives, we devise a decision module
based on Neural Networks that takes in the match percentage from the
chosen pattern recognition algorithms, and then decides the
recognition result based on those match values. Our approach is
evaluated on the Amsterdam Library of Object Images collection, a
large collection of object images containing 1000 objects recorded
under various imaging circumstances. The experimental results
clearly demonstrate that our approach significantly outperforms the
state-of-the-art methods for combining color, shape and texture
features. The developed method is shown to be effective under a wide
variety of imaging conditions. Finally, we employ empirical
evaluation to evaluate our approach with the aid of an accuracy
estimation method, such as k-fold cross validation.
Key words: Object Recognition, Feature Extraction, Support Vector Machine (SVM), Discriminant Canonical Correlation (DCC), Locality Preserving Projections (LPP), Neural Network (NN).
- Koprowski R., Wrobel Z., Zieleznik W.:
Ultrasound Image Analysis in Hashimoto's Disease.
MGV vol. 19, no. 4, 2010, pp. 429-437.
- The paper presents the diagnostics of parenchyma echogenicity and
organ dimensions in thyroid examinations in case of the Hashimoto's
disease using image processing methods. In case of discovering
focal changes within the thyroid, a method for their pathology
evaluation is suggested. The detector proposed operates fully
automatically; using the information on the image texture it detects
an artery in the image, which plays the role of a reference point,
and based on it -- detects the area of interest. }
Key words: USG,
image processing, Hashimoto.
- Sharma A., Sharma R.K., Kumar R.:
HMM-based Online Handwritten Gurmukhi Character
MGV vol. 19, no. 4, 2010,
- This paper presents a hidden Markov model-based online
handwritten character recognition for Gurmukhi script. We discuss a
procedure to develop a hidden Markov model database in order to
recognize Gurmukhi characters. A test with 60 handwritten samples,
where each sample includes 41 Gurmukhi characters, shows a 91.95\%
recognition rate, and an average recognition speed of 0.112 seconds
per stroke. The hidden Markov model database has been developed in
XML using 5330 Gurmukhi characters. This work shall be useful to
implement a hidden Markov model in online handwriting recognition
and its software development.
Online handwriting recognition, Preprocessing, Feature extraction,
Hidden Markov models.
- Ying~J., Zhang~X.:
A Double-Circle Algorithm for Ore Classification.
MGV vol. 19, no. 4, 2010, pp. 451-462.
- This paper proposes a double-circle algorithm to classify ore stockpiles
according to their particle size distribution. The algorithm is particularly
suitable for yard automation systems in large iron and steel works,
since its result can be used directly as a reliable basis for stackers
and reclaimers controller. The paper explains the concept as well
as related method, which consists of four steps to detect ore granularity
and classes. A series of experiments in industrial environments
proved that this novel algorithm improves
the reliability of ore classifiers, compared with the classic ones.
Key words: Particle Size Distribution, Ore Classification, Morphologic Granulation,
Planar Subdivision, Double-circle, Computer Vision.
- Huang F., Torii A., Klette R.:
Geometries of Panoramic Images and 3D Vision.
MGV vol. 19, no. 4, 2010, pp. 463-477.
- Over the recent years, the
image sensor technology has provided
tools for wide-angle and high-resolution 3D recording, analysis and
modeling of static or dynamic scenes, ranging from small objects,
such as artifacts in a museum, to large-scale 3D models of castles
or 3D city maps, also allowing real time 3D data acquisition from a
moving platform, e.g. in vision-based driver assistance. More recently,
due to the rapidly evolving and improving stereoscopic display technology,
many of these panoramic image applications have started to contribute to stereo
visualization, thus increasing realistic and immersive appearances. This
paper introduces a methodology for stereo panorama acquisition and provides
detailed technologies of mapping between different forms of panoramic images.
Image examples illustrate the potential for projects in arts, science and technology.
Key words: Panoramic imaging, stereo panorama
imaging, omnidirectional viewing.
New Books Notes
MGV vol. 19, no. 4, 2010, pp. 479-480.
Katarzyna Stapor: Pattern classification methods
in computer vision (in Polish).
Published by Wydawnictwo Naukowe
PWN, Warszawa 2011.
New Books Notes
MGV vol. 19, no. 4, 2010, pp. 481-482.
Wojciech S. Mokrzycki:
Introduction to Visual Information Processing. I. Perception, Acquisition, Visualization (in Polish).
Published by AOW EXIT, Warsaw 2010.