Skeleton Representation Research Articles

A graph-based method for fitting planar B-spline curves with intersections

Abstract The problem of fitting B-spline curves to planar point clouds is studied in this paper. A novel method is proposed to deal with the most challenging case where multiple intersecting curves or curves with self-intersection are necessary for shape representation. A method based on Delauney Triangulation of data points is developed to identify connected components which is also capable of removing outliers. A skeleton representation is utilized to represent the topological structure which is further used to create a weighted graph for deciding the merging of curve segments. Different to existing approaches which utilize local shape information near intersections, our method considers shape characteristics of curve segments in a larger scope and is thus capable of giving more satisfactory results. By fitting each group of data points with a B-spline curve, we solve the problems of curve structure reconstruction from point clouds, as well as the vectorization of simple line drawing images by drawing lines reconstruction.

LOCALIZATION AND RECOGNITION OF DYNAMIC HAND GESTURES BASED ON HIERARCHY OF MANIFOLD CLASSIFIERS

Abstract. Generally, the dynamic hand gestures are captured in continuous video sequences, and a gesture recognition system ought to extract the robust features automatically. This task involves the highly challenging spatio-temporal variations of dynamic hand gestures. The proposed method is based on two-level manifold classifiers including the trajectory classifiers in any time instants and the posture classifiers of sub-gestures in selected time instants. The trajectory classifiers contain skin detector, normalized skeleton representation of one or two hands, and motion history representing by motion vectors normalized through predetermined directions (8 and 16 in our case). Each dynamic gesture is separated into a set of sub-gestures in order to predict a trajectory and remove those samples of gestures, which do not satisfy to current trajectory. The posture classifiers involve the normalized skeleton representation of palm and fingers and relative finger positions using fingertips. The min-max criterion is used for trajectory recognition, and the decision tree technique was applied for posture recognition of sub-gestures. For experiments, a dataset “Multi-modal Gesture Recognition Challenge 2013: Dataset and Results” including 393 dynamic hand-gestures was chosen. The proposed method yielded 84–91% recognition accuracy, in average, for restricted set of dynamic gestures.

AUTOMATIC DETECTION AND CLASSIFICATION OF RETINAL VASCULAR LANDMARKS

The main contribution of this paper is introducing a method to distinguish between different landmarks of the retina: bifurcations and crossings. The methodology may help in differentiating between arteries and veins and is useful in identifying diseases and other special pathologies, too. The method does not need any special skills, thus it can be assimilated to an automatic way for pinpointing landmarks; moreover it gives good responses for very small vessels. A skeletonized representation, taken out from the segmented binary image (obtained through a preprocessing step), is used to identify pixels with three or more neighbors. Then, the junction points are classified into bifurcations or crossovers depending on their geometrical and topological properties such as width, direction and connectivity of the surrounding segments. The proposed approach is applied to the public-domain DRIVE and STARE datasets and compared with the state-of-the-art methods using proper validation parameters. The method was successful in identifying the majority of the landmarks; the average correctly identified bifurcations in both DRIVE and STARE datasets for the recall and precision values are: 95.4% and 87.1% respectively; also for the crossovers, the recall and precision values are: 87.6% and 90.5% respectively; thus outperforming other studies.

A custom image-based analysis tool for quantifying elastin and collagen micro-architecture in the wall of the human aorta from multi-photon microscopy

The aorta possesses a micro-architecture that imparts and supports a high degree of compliance and mechanical strength. Alteration of the quantity and/or arrangement of the main load-bearing components of this micro-architecture – the elastin and collagen fibers – leads to mechanical, and hence functional, changes associated with aortic disease and aging. Therefore, in the future, the ability to rigorously characterize the wall fiber micro-architecture could provide insight into the complicated mechanisms of aortic wall remodeling in aging and disease. Elastin and collagen fibers can be observed using state-of-the-art multi-photon microscopy. Image-analysis algorithms have been effective at characterizing fibrous constructs using various microscopy modalities. The objective of this study was to develop a custom MATLAB-language automated image-based analysis tool to describe multiple parameters of elastin and collagen micro-architecture in human soft fibrous tissue samples using multi-photon microscopy images. Human aortic tissue samples were used to develop the code. The tool smooths, cleans and equalizes fiber intensities in the image before segmenting the fibers into a binary image. The binary image is cleaned and thinned to a fiber skeleton representation of the image. The developed software analyzes the fiber skeleton to obtain intersections, fiber orientation, concentration, porosity, diameter distribution, segment length and tortuosity. In the future, the developed custom image-based analysis tool can be used to describe the micro-architecture of aortic wall samples in a variety of conditions. While this work targeted the aorta, the software has the potential to describe the architecture of other fibrous materials, tube-like networks and connective tissues.

A simple fracture energy prediction method for fiber network based on its morphological features extracted by X-ray tomography

The fracture behavior of a novel porous metal fiber sintered sheet (PMFSS) was predicted using a semi-empirical method combining the knowledge of its morphological characteristics and micro-mechanical responses. The morphological characteristics were systematically summarized based on the analysis of the topologically identical skeleton representation extracted from the X-ray tomography images. The analytical model firstly proposed by Tan et al. [1] was further modified according to the experimental observations from both tensile tests of single fibers and sintered fiber sheets, which built the coupling of single fiber segment and fiber network in terms of fracture energy using a simple prediction method. The efficacy of the prediction model was verified by comparing the predicted results to the experimental measurements. The prediction error that arose at high porosity was analyzed through fiber orientation distribution. Moreover, the tensile fracture process evolving from single fiber segments at micro-scale to the global mechanical performance was investigated.

L 1 -medial skeleton of point cloud

We introduce L 1 - medial skeleton as a curve skeleton representation for 3D point cloud data. The L 1 -median is well-known as a robust global center of an arbitrary set of points. We make the key observation that adapting L 1 -medians locally to a point set representing a 3D shape gives rise to a one-dimensional structure, which can be seen as a localized center of the shape. The primary advantage of our approach is that it does not place strong requirements on the quality of the input point cloud nor on the geometry or topology of the captured shape. We develop a L 1 -medial skeleton construction algorithm, which can be directly applied to an unoriented raw point scan with significant noise, outliers, and large areas of missing data. We demonstrate L 1 -medial skeletons extracted from raw scans of a variety of shapes, including those modeling high-genus 3D objects, plant-like structures, and curve networks.

Analysis of short fibres orientation in steel fibre-reinforced concrete (SFRC) by X-ray tomography

The mechanical properties of fibre composite materials are largely determined by the orientation of fibres within the matrix. Which orientation distribution short fibres follow in different parts of a structural element is still a subject for research and discussions in the scientific community. In this article, we present a modern and advanced method for measuring the orientation of short fibres in steel fibre-reinforced concrete (SFRC) by X-ray microtomography. With this method, a voxel image of the fibres is obtained directly in 3D, and the orientation of each individual fibre is calculated based on a skeletonized representation of this image. Scans of 12 SFRC samples, taken from the central height region of real-size floor slabs, reveal the fibres to be mostly horizontally oriented near the centre of a floor slab and more vertically oriented near the edge; here the alignment with the formwork dominates. The fibre orientation distributions are characterized by several orientation parameters as quantitative measures for the alignment. On the practical side, this method has the potential to be incorporated into the development and production process of SFRC structures to verify how the fibres contribute to capacity.

Morphological shape comparison based on skeleton representations

The concept of a pattern-shape comparison peculiar to the Pyt'ev morphological paradigm is introduced for monotonic morphologies of the Serra type. Morphological skeletons with variable radii (scales) of elements are considered to be analogs of Pyt'ev partitions of a frame into regions with variable brightness. This allows one to construct projections and calculate the morphological coefficients of the correlation.

Automated extraction of three-dimensional cereal plant structures from two-dimensional orthographic images

The authors present a novel approach to automatically extract three-dimensional (3D) structures of cereal plants from 2D orthographic images. The idea of this approach is to use 2D skeletons to represent elongated shapes of plants in 2D images and then to generate 3D plant structures from 2D skeletons. However, existing skeletonisation algorithms generate artefacts because of boundary extremities and noise. Moreover, cereal plant leaves appear as broken segments in images because of leaf twists. In this approach, oriented Gaussian filters are used to obtain ridges from 2D images via guided skeletonisation, so that one can obtain smooth skeleton representation of 2D elongated shapes without redundant branches. For broken leaf segments, a cost function is proposed to identify whether the two segments belong to the same leaf and thus should be connected. For a given leaf tip of a cereal plant, the z-coordinators of two 2D side-view images are same. Based on this, it is easy to identify the same leaf tip in two 2D side-views. Thus, the authors are able to track skeletons of leaves from their tips to the plant roots. Experimental results show that the proposed approach is able to handle various issues such as broken segments and overlapping among plant parts, and is also able to automatically extract 3D structures of cereal plants.

Skeleton pruning by contour approximation and the integer medial axis transform

We present a new shape skeleton pruning algorithm based on contour approximation and the integer medial axis. The algorithm effectively removes unwanted branches, conserves the connectivity of the skeleton and respects the topological properties of the shape. The algorithm is robust to significant boundary noise and to rigid shape transformations, it is fast and easy to implement. High accuracy reconstruction of the shape is possible from the generated skeleton by means of the integer medial axis transform. Our algorithm also produces a vector representation of the skeleton. We compare our algorithm with state-of-the-art techniques for computing stable skeleton representations of shapes including pruning. We test and compare our solution using the MPEG-7 CE Shape-1 Part B dataset looking for skeleton connectivity, complexity, parameter selection, and accuracy/quality of the outcome. The experimental results show that our solution outperforms existing solutions according to these criteria.

Recombinant Skeleton Using Junction Points in Skeleton Based Images

We perform the task of combining two skeleton images and to produce the recombinant skeleton. We propose the recombinant skeleton algorithm to produce the recombinant skeletons. The existing skeleton representation has been taken and the merge vertex detection algorithm was used before applying the recombinant skeleton algorithm. We can design and apply this recombinant skeleton in motion detection, image matching, tracking, panorama stitching, 3D modeling and object recognition. We can generate or manufacture the true real time object from the recombinant skeleton produced. The proposed method utilize local search algorithm for junction validation. Our frame work suggests the range of possibility in getting the recombinant skeleton. The boundary is essential for any transformation hence the bamboo skeleton algorithm is deployed for computing the boundary and for storing the skeleton together with the boundary. Thus our representation is skeleton with border or outline. From this new skeleton representation the proposed recombinant is achieved.

Visualizing High‐Order Symmetric Tensor Field Structure with Differential Operators

The challenge of tensor field visualization is to provide simple and comprehensible representations of data which vary both directionally and spatially. We explore the use of differential operators to extract features from tensor fields. These features can be used to generate skeleton representations of the data that accurately characterize the global field structure. Previously, vector field operators such as gradient, divergence, and curl have previously been used to visualize of flow fields. In this paper, we use generalizations of these operators to locate and classify tensor field degenerate points and to partition the field into regions of homogeneous behavior. We describe the implementation of our feature extraction and demonstrate our new techniques on synthetic data sets of order 2, 3 and 4.

A similarity-based approach for shape classification using Aslan skeletons

Shape skeletons are commonly used in generic shape recognition as they capture part hierarchy, providing a structural representation of shapes. However, their potential for shape classification has not been investigated much. In this study, we present a similarity-based approach for classifying 2D shapes based on their Aslan skeletons ( Aslan and Tari, 2005; Aslan et al., 2008). The coarse structure of this skeleton representation allows us to represent each shape category in the form of a reduced set of prototypical trees, offering an alternative solution to the problem of selecting the best representative examples. The ensemble of these category prototypes is then used to form a similarity-based representation space in which the similarities between a given shape and the prototypes are computed using a tree edit distance algorithm, and support vector machine (SVM) classifiers are used to predict the category membership of the shape based on computed similarities.

INTERPRETATION OF AMBIGUOUS ZONE TO IMPROVE THINNING RESULTS OF HANDWRITTEN CHINESE CHARACTERS

The skeleton representation of characters is a fundamental step to handwriting recognition, but traditional skeletonization algorithms always produce unwanted artifacts or pattern distortions at regions with intersections or junctions of strokes. In this paper, we propose a novel method to eliminate these unreliable skeleton segments and improve the skeletonization of handwritten Chinese characters on the basis of ambiguous zone interpretation. This method consists of two main phases. In the first phase, the parts of characters which contain the distortions of skeleton, called ambiguous zones, are detected. Instead of exploiting any corner or dominant point detection, a set of feature points from the original skeleton and the contour information around them are manipulated in our approach. In the phase of interpretation, the continuity of skeleton segments of substrokes is estimated based on the minimum curvature variation criterion, and compensations are made to fix interstices of terminated skeleton segments. Finally, the distorted parts of characters are reconstructed by applying a cubic B-spline interpolation. Experimental results show that the proposed method can detect ambiguous zones with arbitrary shapes accurately, and produce skeletons that are close to human perceptions.

Finite volume distance field and its application to medial axis transforms

AbstractAccurate and efficient computation of the nearest wall distance d (or level set) is important for many areas of computational science/engineering. Differential equation‐based distance/level set algorithms, such as the hyperbolic‐natured Eikonal equation, have demonstrated valuable computational efficiency. Here, in the context, as an ‘auxiliary’ equation to the main flow equations, the Eikonal equation is solved efficiently with two different finite volume approaches (the cell‐vertex and cell‐centered). The application of the distance solution is studied for various geometries. Moreover, a procedure using the differential field to obtain the medial axis transform (MAT) for different geometries is presented. The latter provides a skeleton representation of geometric models that has many useful analysis properties. As an alternative to other methods, the current d‐MAT procedure bypasses difficulties that are usually encountered by pure geometric methods (e.g. the Voronoi approach), especially in three dimensions, and provides better accuracy than pure thinning methods. It is also shown that the d‐MAT approach provides the potential to sculpt/control the MAT form for specialized solution purposes. Copyright © 2009 John Wiley & Sons, Ltd.

Processing of fluorescence angiograms for the quantification of vascular effects induced by anti-angiogenic agents in the CAM model

The chicken embryo chorioallantoic membrane (CAM) is widely used as an in vivo model to study the vascular effects of angiogenesis modulating agents. The main goal of the present study was to develop and validate a quantitative method to characterize time-dependent changes mainly in the capillary network of the CAM. To that end, the CAM capillaries were monitored in ovo, between days 7 and 13 of embryo development, using an epi-fluorescence microscope equipped with a sensitive camera following the intravenous injection of a fluorescent agent. We present a method by which the fluorescence angiograms of the CAM vasculature were recorded and analyzed by a multistep mathematical procedure to obtain a skeleton representation of the vessels and capillaries. From this skeleton descriptors were extracted, including the number of branching points/mm 2, the mean area of the vessel network meshes, and the mean of the 3rd quartile of the mesh area histogram. A qualitative visual assessment of the vasculature based on the number and the size of the avascular zones was obtained for comparison. To illustrate this approach, the activity of a neutralizing anti-vascular endothelial growth factor antibody, Avastin ® (Bevacizumab), was then quantified. Blood vessel growth inhibition and changes in the architecture of the capillary plexus after topical application of this drug on the CAM surface were monitored using the three descriptors mentioned above.

Local Thickness and Anisotropy Approaches to Characterize Pore Size Distribution of Three-Dimensional Porous Networks

Two image-analysis approaches for pore size distribution (PSD) of porous media are proposed. The methods are based on the skeleton representation of a porous object. One approach gives the local thickness of the pore object to represent the pore size corresponding to a lower limit of PSD. The other gives the pore size taking into account the anisotropy of pore object and corresponds to an upper limit of PSD. These two approaches can be incorporated into a computer program without computationally intensive and complex mathematical operations. In this study, these two approaches are applied to a two-dimensional (2D) synthetic image and 3D natural images of tissue scaffolds with various porosities and tortuosities. The scaffolds were prepared by removing the water-soluble poly(ethylene oxide) (PEO) component of the polycaprolactone (PCL)/PEO blend, leaving a porous PCL scaffold. Extracting quantitative PSD information for materials with an interconnected porous network rather than discrete voids (such as tissue scaffolds) is inevitably subjective without a universally accepted definition of "pore size." Therefore, the proposed lower and upper limits of PSD can come into play when considering mass transfer and scaffold surface area for cell-matrix interaction.

Dissimilarity between two skeletal trees in a context

Skeletal trees are commonly used in order to express geometric properties of the shape. Accordingly, tree-edit distance is used to compute a dissimilarity between two given shapes. We present a new tree-edit based shape matching method which uses a recent coarse skeleton representation. The coarse skeleton representation allows us to represent both shapes and shape categories in the form of depth-1 trees. Consequently, we can easily integrate the influence of the categories into shape dissimilarity measurements. The new dissimilarity measure gives a better within group versus between group separation, and it mimics the asymmetric nature of human similarity judgements.

Computer Vision Approach for Ultrasound Doppler Angle Estimation

Doppler ultrasound is an important noninvasive diagnostic tool for cardiovascular diseases. Modern ultrasound imaging systems utilize spectral Doppler techniques for quantitative evaluation of blood flow velocities, and these measurements play a crucial rule in the diagnosis and grading of arterial stenosis. One drawback of Doppler-based blood flow quantification is that the operator has to manually specify the angle between the Doppler ultrasound beam and the vessel orientation, which is called the Doppler angle, in order to calculate flow velocities. In this paper, we will describe a computer vision approach to automate the Doppler angle estimation. Our approach starts with the segmentation of blood vessels in ultrasound color Doppler images. The segmentation step is followed by an estimation technique for the Doppler angle based on a skeleton representation of the segmented vessel. We conducted preliminary clinical experiments to evaluate the agreement between the expert operator's angle specification and the new automated method. Statistical regression analysis showed strong agreement between the manual and automated methods. We hypothesize that the automation of the Doppler angle will enhance the workflow of the ultrasound Doppler exam and achieve more standardized clinical outcome.

Topologically Clean Distance Fields

Analysis of the results obtained from material simulations is important in the physical sciences. Our research was motivated by the need to investigate the properties of a simulated porous solid as it is hit by a projectile. This paper describes two techniques for the generation of distance fields containing a minimal number of topological features, and we use them to identify features of the material. We focus on distance fields defined on a volumetric domain considering the distance to a given surface embedded within the domain. Topological features of the field are characterized by its critical points. Our first method begins with a distance field that is computed using a standard approach, and simplifies this field using ideas from Morse theory. We present a procedure for identifying and extracting a feature set through analysis of the MS complex, and apply it to find the invariants in the clean distance field. Our second method proceeds by advancing a front, beginning at the surface, and locally controlling the creation of new critical points. We demonstrate the value of topologically clean distance fields for the analysis of filament structures in porous solids. Our methods produce a curved skeleton representation of the filaments that helps material scientists to perform a detailed qualitative and quantitative analysis of pores, and hence infer important material properties. Furthermore, we provide a set of criteria for finding the "difference" between two skeletal structures, and use this to examine how the structure of the porous solid changes over several timesteps in the simulation of the particle impact.