• No. 1/2: Special Issue on Novel Methods in Computer-Aided Biomedical Image Processing and Interpretation. Part II
  • No. 3
  • No. 4

Machine GRAPHICS & VISION, Vol. 17 (2008), No. 1/2:

Special Issue on Novel Methods in Computer-Aided Biomedical Image Processing and Interpretation. Part II
Special Issue Editor: Juliusz L. Kulikowski.

Kulikowski J.L.:
Introduction to the special issue of MG&V.
MGV vol. 17, no. 1/2, 2008, pp. 3-4.

Orkisz M., Flórez Valencia L., Hernández Hoyos M.:
Models, algorithms and applications in vascular image segmentation.
MGV vol. 17, no. 1/2, 2008, pp. 5-33.

A synthesis of the authors' projects in the field of 3D vascular image processing in the last decadeis provided. This work was motivated by the following applications: display improvement, extraction of geometrical measurements, acquisition optimization, stent-pose planning, phantom generation, blood-flow simulations. The methods are often dependent on the imaging modality and/or on the anatomic region. They involve both: low-level models of intensity patterns and profiles, and higher-level models of cylindrical shapes. Amongst the various algorithms used, recursive tracking and fast-marching level-sets are emphasized. Critical analysis of each model and algorithm is carried out. Problems that remain open, and perspectives associated with the progress of the image acquisition techniques, are listed.
Key words:inertia moments, Hessian, eigen-analysis, centerlines, active contours, deformable models, generalized cylinders, simplex meshes, tracking, fast-marching level sets.

Garcia E., Seron F., Baldassarri S.:
The challenge of hexahedral meshing of arterial geometry.
MGV vol. 17, no. 1/2, 2008, pp. 35-55.

This paper describes the process of generating a hexahedral mesh of the arterial geometry. From the finite element meshing point of view, arterial geometry may be regarded as a volume assembly in which every segment can be meshed separately except in the shared surfaces. The arterial assembly is made up of three subvolumes: arterial wall, artery lumen and atherosclerotic plaque. A three-dimensional geometric model of each arterial segment has been reconstructed using intravascular ultrasound images (IVUS) and biplane angiographies. Generation of hexahedral meshes for biological models with different physical characteristics usually requires the use of different meshing algorithms for each region. Vessel bifurcations have been modeled by joining the surfaces of the reconstructed segments, using a technique based on NURBS. Therefore, this paper describes the combination of decomposition and meshing techniques required to meet the challenge of generating hexahedral elements for arterial models. A variety of verification algorithms have been used in order to calculate several algebraic quality metrics and assess the quality of the finite element meshes generated.
Key words: mesh generation, hexahedra, arterial bifurcations, atherosclerosis

Jelinek H. F.:
Automated vessel segmentation of 35mm colour non-mydriatic images in a community health screening project.
MGV vol. 17, no. 1/2, 2008, pp. 57-68.

Proliferative diabetic retinopathy is a complication of diabetes that can eventually lead to blindness. Early identification of this complication reduces the risk of blindness so that timely treatment can be initiated. In rural and remote regions, widespread population screening is practically impossible due to the lack of ophthalmologists and the cost associated with rural visits by specialists. Several methods for vessel segmentation have been discussed in the literature, but none have used non-mydriatic colour images obtained from community screening initiatives. Rural screening clinics currently use either 35mm or Polaroid photography. In addition, the quality of the images is often much lower. Scanning images at 300dpi provides very low resolution images which combined with the low quality requires a robust algorithm to identify vessels with high accuracy. Visual inspection by an ophthalmologist judged 46 images (88%) to represent an acceptable level of segmentation. Despite the low resolution and quality of images, the Gabor wavelet provided vessel segmentation results that were usable in rural community screening projects and in some cases identified vessels obscured by haemorrhages better than the expert observer.
Key words: vessel segmentation, wavelet transform, non-mydriatic, diabetic retinopathy.

Goszczynska H.:
A method for densitometric analysis of moving object tracking in medical images.
MGV vol. 17, no. 1/2, 2008, pp. 69-90.

The aim of this work is to develop a method of automatic data collection for coronary blood flow estimation based on computer analysis of angiographic image sequences. In the methods of coronary flow measurements, apart from problems involving densitometric analysis, there is also the problem of cyclic movement of the measurement field, i.e. of an artery segment or part of the myocardium. The system of automatic artery segment tracking presented in this paper does not only reduce the "manual effort" of the operator to establish the region of interest in the frame sequence, but also makes it possible to plot a densitometric curve with the time resolution equal to that of the frame sequence. The algorithm thus implemented, based on a template matching method, makes it also possible to trace the results of automatic detection of some characteristic points within the structure of arteries and to correct any faulty matching. The incorporation of the movement trajectories obtained allows us to trace the movement of the part of the myocardium close to the characteristic points of the artery with the aim of estimating the degree of myocardium perfusion. This relatively simple algorithm is acceptable for routine clinical testing due to the short time of frame sequence analysis (few minutes) and its relatively small error (the maximum estimated error of the automatic analysis is less than 11%). The comparative analysis of the results obtained for the template matching algorithm based on several criteria of similarity failed to establish any specific criterion with regard to acceleration or matching accuracy.
Key words: moving object tracking, similarity measures, image registration, densitometric measurements.

Cha S.-H., Gargano M. L., Chang S., Quintas L. V., Wahl E. M.:
Estimating a vascular network growth using random graphs.
MGV vol. 17, no. 1/2, 2008, pp. 91-103.

Vascular networks develop by way of angiogenesis, a growth process that involves the biological mechanisms of vessel sprouting (budding) and splitting (intussusception). Graph theory is excellently suited to model vascular networks and to analyze their properties (invariants). In particular, a random graph process model can simulate the development of a vascular network that has been modeled using graph theory. The renal glomerulus is one example of such a vascular network. Here the correlation between the invariants of this vascular network modeled as a graph and the mechanisms of the network growth using a random graph process are studied. It is proposed that the relative frequencies of sprouting and splitting during the growth of a given renal glomerulus can be estimated by the invariants (root distance, radius, and diameter) of the graph representing the renal glomerulus network. Experimental evidence has been given to support this conjecture.
Key words: Renal glomerular network, random graph process, graph invariants, pattern matching.

Klette G.:
Skeletal curves of 3D astrocyte samples.
MGV vol. 17, no. 1/2, 2008, pp. 105-129.

The paper discusses the concept of simple (and non-simple) elements for the generation of topologic skeletons, their transformation into abstract curve graphs, and the analysis of such graphs.
The definition of a branching index of a point on a curve is fundamental in curve theory (in Euclidean space), and leads to important subjects of curve analysis. This paper derives analogous notions, such as branching index, branch element, and junction, for digital curves, which allow us to introduce new concepts for analyzing complex digital curves in a 3D space. The paper provides new theoretical insights, and also discusses an application project (the description of astrocytes in 3D confocal images of human brain tissue).
This work was originally initiated by a particular research project at the Medical School of The University of Auckland. Medical experts developed the hypothesis that features of astrocytes in confocal volume scans are useful for defining states between normal and abnormal tissue. The calculation of skeletal curves, as proposed and studied in this paper, provides a valuable tool for calculating such features.
Key words: shape analysis, skeletonization, thinning, astrocytes.

Pietka B. D., Dulewicz A., Jaszczak P.:
Removing artefacts from microscopic images of cytological smears.
MGV vol. 17, no. 1/2, 2008, pp. 131-152.

This paper gives an overview of authors' attempts to design a computer-assisted urine smear screening system, focusing on the nastiest issues hampering its successful practical implementation. There are many valuable works concerning more or less sophisticated image processing and data mining algorithms which are capable of automatically detecting pathological morphology of cytological objects and distinguishing them from normal ones. Unfortunately, most of the attempts to implement those smart ideas in real world are likely to fail because of one but fundamental obstacle – artefacts. If not properly identified and removed from the analysis, they tend to generate so many false-positive warnings that the automated support is going to be useless because of its dramatically low specificity. Our paper addresses this neglected problem, trying to point out some general rules and implementation details that should be followed to reduce the influence of artefacts on overall system performance.
Key words: computer-aided digital cytology, urologic oncology, microscopic smears screening systems, urinary bladder cancer.

Marciniak A., Nieczkowski T., Obuchowicz A.:
Color homogram for segmentation of fine needle biopsy images.
MGV vol. 17, no. 1/2, 2008, pp. 153-165.

In this paper, a new weighted clustering algorithm for image segmentation in cytopathology is introduced. The weights incorporating spatial information into pixel-based segmentation are computed with use of a color homogram. The effectiveness of the proposed solution is evaluated on microscopic fine needle biopsy (FNB) images. The results of the classical fuzzy c-means algorithm and its weighted modification are compared.
Key words: homogeneity, color image segmentation, clustering, fine needle biopsy, fuzzy logic.

Witkowski L.:
A computer system for sperm cells motility evaluation.
MGV vol. 17, no. 1/2, 2008, pp. 167-186.

The article describes the first module of a computer system for human sperm assessment. Different image acquisition methods were described and optical microscope with bright optic was selected as the most appropriate one. Image enhancement and processing methods were selected and customized to achieve effective sperm cells recognition and tracking. Trajectories of sperm cells and their movement parameters were found, calculated and classified in compliance with WHO standards.
Key words: image processing, image recognition, sequence of images, microscopy techniques, WHO.

Ramoser H.:
Leukocyte segmentation and SVM classification in blood smear images.
MGV vol. 17, no. 1/2, 2008, pp. 187-200.

Automated leukocyte detection, segmentation, and classification is an important task in clinical diagnosis. In this paper we present an approach to leukocyte cytoplasm and nucleus segmentation that is robust with respect to image quality and cell appearance. Cell properties are described by a set of statistical color and shape features. Pairwise coupling of SVM classification results is used to determine cell type probabilities. Evaluation of the method on a set of 1166 images containing 13 different cell types has resulted in 95% correctly segmented cells and a classification accuracy of 88% (at 20% reject rate).
Key words: cytometry, feature extraction, multiclass SVM.

Koprowski R., Wrobel Z.:
Analysis of the inclination of elongated biological objects - microtubules.
MGV vol. 17, no. 1/2, 2008, pp. 201-215.

In the paper we propose a new method for determining the inclination angle of microtubules. The method allows us to obtaining an angular inclination histogram, taking into account the area of microtubules rather than their number only. We present two kinds of microtubule approximation: global and local one. In the global approach microtubules are approximated with one or more straight lines, whereas the local approach defines a direction field for each pixel.
Key words: microtubules, geometric approximation, field of direction, inclination.

Machine GRAPHICS & VISION, Vol. 17 (2008), No. 3:

Bojar K., Nieniewski M.:
Analysis of temporal variations of dynamic textures by means of the SGLDM with application to EIT solar images.
MGV vol. 17, no. 3, 2008, pp. 219-247.

The paper discusses applicability of texture analysis methods to Extreme-Ultraviolet Imaging Telescope (EIT) images obtained from the SOHO mission, and in particular analyses of the solar texture by means of the Spatial Gray Level Dependence Matrices (SGLDMs). Based on the observations, a theoretical parametric model is proposed for the SGLDM. Subsequently, the Haralick features are computed directly for the time sequence of preprocessed images, as well as and for the proposed model of the SGLDM. Time evolution of model parameters as well as measures of goodness fit of the model to image data for the period of one year is presented. The results obtained confirm the correctness of the model. In view of the simplicity of the model, it gives remarkably good results. A new method for solar limb extraction is described in the Appendix. The method is superior to those presented in the literature since it uses no external parameters, such as thresholds.
Key words: texture analysis, texture modelling, SGLDM, solar images, EIT images.

Rodríguez T.:
Shadow removal for robust vehicle detection.
MGV vol. 17, no. 3, 2008, pp. 249-266.

Shadow removal is a critical process for any computer vision based vehicle detection system. Traditional shadow removal methods are not sufficiently robust since they frequently remove parts of the vehicles together with the shadows. For robust shadow identification and removal, shadow information must be compiled at every step of the computer vision chain from the time the shadow enters the scene until it finally disappears. In this paper we present a shadow removal algorithm that preserves the compactness of the vehicles' object masks while allowing for dealing with changing illumination conditions, long and broken shadows and multiple shadows at night. The method is based on a complex shadow model which includes, among others, information regarding: luminance and chromaticity, morphology, dynamics and spatial relations analysis.
Key words: computer vision, real-time, traffic monitoring, shadow removal.

Rozen T., Boryczko K., Alda W.:
A GPU-based method for approximate real-time fluid flow simulation.
MGV vol. 17, no. 3, 2008, pp. 267-278.

Fluid flow can be realistically simulated by physical models. We present a method for simplifying the Navier-Stokes equations by relaxing the incompressibility constraint. Our method allows for low-cost real-time simulation of two-dimensional fluid flow with accuracy sufficient for computer graphics. The implementation takes advantage of recent programmable floating-point graphics hardware, which performs all the necessary computations.
Key words: fluid dynamics, graphics hardware, physically-based simulation, pseudo-compressibility method.

Korohoda P., Dabrowski A.:
Generalized convolution for extraction of image features in the primary domain.
MGV vol. 17, no. 3, 2008, pp. 279-297.

In this paper, a class of techniques for flexible extraction of image features is proposed. These techniques are based on the convolution type filtering in the primary domain. The mentioned flexibility results from the fact that any discrete transform suitable for the analysis of desired image features (as, e.g., the Karhunen-Loeve transform) may be designed and the filter with the assumed transform-domain properties may be applied straightforwardly in the image (i.e., primary) domain. The proposed solution creates an entirely new approach to image filtering. After recalling the concept referred to by the authors as generalized convolution and extending it to the 2-dimensional case, the theoretical results are illustrated with several examples based on filtering of the test image with filters designed in the Haar, Hadamard and the DCT II domains. Finally, it is explained how the proposed approach indicates several possible ways for further developments towards the design of image-and-feature based tools.
Key words: generalized convolution, discrete transforms, image decomposition, image filtering.

Hati S.:
Estimation of pose parameters from a set of least square objective functions.
MGV vol. 17, no. 3, 2008, pp. 299-312.

In this paper, we have attempted to solve the pose estimation problem for a 3-dimensional object by independently estimating the pose parameters through the minimization of a set of objective, functions, using Gauss approximation techniques for least squares optimization. In our implementation, the 3-D object is assumed to have three degrees of freedom on a flat surface, which is typical of automated visual inspection applications. However, the solution can be also extended to greater degrees of freedom. We have is shown that the pose can be estimated by only considering the x-coordinates of the known vertices in the projected space, but the same is not true if we consider the y-coordinates alone. We propose a set of modified objective functions from which it is possible to find the pose parameters. The parameters have been determined in noisy conditions under 20-dB and 40-dB SNR values and the robustness of the estimators is confirmed.
Key words: pose estimation, camera calibration, least square estimation, robust estimator.

Zhang Y., Sung E.:
EDFCES: A new example-driven 3D face construction and editing system.
MGV vol. 17, no. 3, 2008, pp. 313-346.

This paper presents an automatic runtime system for generating varied, realistic face models by synthesizing a global face shape and local facial features according to intuitive, high-level control parameters. Our method takes as examples 3D face scans in order to exploit the parameter-to-geometry correlations present in real faces. In order to establish the correspondences among the scanned models, we use a three-step model fitting approach to conform a generic head mesh onto each scanned model. We transform the obtained data sets of global face shapes and local feature shapes into vector space representations by applying a principal component analysis (PCA). We compute a set of face anthropometric measurements to parameterize the exemplary shapes in the measurement spaces. Using PCA coefficients as a compact shape representation, we approach the shape synthesis problem by forming scattered data interpolation functions designed to generate the desired face shape by taking anthropometric parameters as input. At runtime, the interpolation functions are evaluated for the input parameter values to produce new face geometries at an interactive rate. The correspondence among all exemplary face textures is obtained by parameterizing the 3D generic mesh over a 2D image domain. The new feature texture with the desired attributes is synthesized by interpolating the example textures. The resulting system is intuitive to control and fine-grained. We demonstrate our method by applying different parameters to generate a wide range of face models.
Key words: face modeling, facial features, anthropometry, interpolation, PCA, model fitting, 3D scanned data.


Machine GRAPHICS & VISION, Vol. 17 (2008), No. 4:

Sitnik R., Karaszewski M.:
Optimized Point Cloud Triangulation for 3D Scanning Systems.
MGV vol. 17, no. 4, 2008, pp. 349-371.

This article deals with a new Mesh3D algorithm used for triangulation of point clouds obtained from real 3D objects via optical scanning systems. The algorithm is characterized by a high level of automation. Triangle meshes obtained by using this algorithm are of a better quality than those created through the application of popular algorithms. Procedures used for areas with large shape gradients allow us to obtain a higher accuracy of a geometric object representation. It is also a very ef\mbox{}f\mbox{}icient method even when the procedure of point clouds smoothing is applied initially. The implementation of this kind of a complex mesh quality assessment system enables us to improve the smoothness and continuity of the mesh, which allows us to obtain more realistic and visually better reconstructions of objects.
Key words: triangulation, 3D scanning, mesh, point cloud.

Titov V.S., Degtyarev S.V., Truphanov M.I.:
The Calibration Method for Stereoscopic Vision System.
MGV vol. 17, no. 4, 2008, pp. 373-387.

Stereoscopic vision systems are used not only in visual design computing but also in many other applications. In stereoscopic vision, an important property is the accuracy of three-dimensional reconstruction. This property depends considerably on the quality of the vision system calibration. The known solutions to the calibration problem are based on determining calibration parameters from an image of a special pattern. Our method allows calibration of the stereoscopic vision without such a special pattern. The calibration includes the following basic stages: selection of an object used in the calibration (one a priori unknown object instead of a special model object) which is chosen from a set of objects existing in a working scene, calibration of the angles between two cameras which are part of the stereoscopic vision system, and calibration of the distances between two cameras. The calibration parameters are calculated with the aid of images of calibration objects. This approach allows us to perform an adaptive calibration of the vision system (automatic calibration is performed from time to time when necessary) because this process does not require placing of the special calibration object in the working scene, it does not interrupt execution of main function of the vision system and increases the calibration accuracy since possible errors which may be introduced during the placement of the calibration object do not affect the result of the calibration.
Key words: stereo vision, calibration, fuzzy logic, automatic, accuracy.

El-Khamy S.E., Hadhoud M.M., Dessouky M.I., Salam B.M., Abd El-Samie F.E.:
A Wavelet based Entropic Approach To High-Resolution Reconstruction of Images.
MGV vol. 17, no. 4, 2008, pp. 389-402.

In multi-channel imaging, when multiple limited resolution sensors are used, we need to extract a high resolution (HR) image from the available low resolution (LR) observations. In this paper, an entropic approach to the HR reconstruction of images is developed. The suggested approach is based on breaking the HR reconstruction problem into three consecutive steps; the multi channel restoration step, the image fusion step and the image interpolation step. In both the first and the last steps, the maximum entropy concept is used to obtain an output HR image with the maximum amount of information. The image fusion step is based on the wavelet approach. The objective of this step is to integrate the data obtained from each observation into a single image, which is then interpolated to give an HR image. The paper explains the necessary approximations required to reduce the computational complexity of each step. The suggested approach has proved to be a computationally efficient for HR reconstruction of images.
Key words: maximum entropy restoration, image fusion, multi-channel restoration, maximum entropy interpolation, toplitz to circulant approximation.

Shahbe M.D., Hati S.:
Decision Fusion Based on Voting Scheme for IR and Visible Face Recognition.
MGV vol. 17, no. 4, 2008, pp. 403-424.

In this paper we present an evaluation study of decision fusion strategies for infrared (IR) and visible face recognition. Several decision fusion methods based on a voting scheme (minimization, product and averaging) are discussed, and experiments for various conditions of probe and gallery sets are performed on two databases with paired IR and visible face imageries. The Eigenfaces and Fisherface classification techniques are used to extract the face features, and the performance of fusion methods on both classification approaches is discussed.
Key words: face recognition, IR, visible, decision fusion, eigenfaces, Fisherfaces.

Mikrut Z., Kubiak A.:
Recognition of Wheat Grain Quality Using Log-Hough Representation and Neural Networks.
MGV vol. 17, no. 4, 2009, pp. 425-449.

Assessment of raw product quality constitutes one of the most important issues in the agricultural sectors of food production, processing and storage. In wheat grain quality assessment, the evaluation of the percentage of broken grains in a single variety sample is one of the most important criteria. In the present work, we propose a solution based on a computer vision system and neural networks. An algorithm which performs normalization of the size and rotation angle of a single grain image in the log-polar space is developed. The grain edge image is subsequently transformed to the accumulative log-Hough space and projected onto the coordinate system axes. The resulting representation undergoes classification and variety discrimination with the use of the Kohonen Self Organizing Map. The effectiveness of this representation has been verified with the use of a backpropagation neural network and the k-Nearest Neighbors method. The average classification rate within a single wheat variety exceeds 97%, which qualifies the method for practical applications.
Key words: log-polar transform, log-Hough transform, wheat grain, shape recognition.

Koprowski R., Wrobel Z., Kucypera K.:
EDFCES: 3D Modeling of the Growth and Division of Shoot Apex Meristem.
MGV vol. 17, no. 4, 2008, pp. 451-459.

The paper presents an approach to three-dimensional visualization of idealized growth and division of the shoot apex meristem of a plant root based on simplification of cells to the form of non-complex polyhedrons rendered by a graphic library. The vertices of polyhedrons are processed by a growth algorithm, which determines their position in space and successively increases their size until the division criterion is fulfilled and division is triggered. The proposed method allows simulation of the aforementioned processes and observation of their characteristics through animation of a single cell or large sets of cells.
Key words: three-dimensional visualization, turgor pressure, growth tensor, meristem, cell growth simulator.

Call for Papers – a special issue on Image Databases

Reviewers' index

Authors' index

First Negative Selection Ratio, Final Acceptance Ratio

Contents of volume 17, 2008