Statistical Shape Models Using Elastic-String Representations

Size: px
Start display at page:

Download "Statistical Shape Models Using Elastic-String Representations"

Transcription

1 Statistical Shape Models Using Elastic-String Representations Anuj Srivastava, Aastha Jain, Shantanu Joshi,andDavidKaziska Florida State University, Tallahassee, FL, USA Indian Institute of Technology, N. Delhi, India Air Force Institute of Technology, Dayton, OH, USA Abstract. To develop statistical models for shapes, we utilize an elastic string representation where curves (denoting shapes) can bend and locally stretch (or compress) to optimally match each other, resulting in geodesic paths on shape spaces. We develop statistical models for capturing variability under the elastic-string representation. The basic idea is to project observed shapes onto the tangent spaces at sample means, and use finite-dimensional approximations of these projections to impose probability models. We investigate the use of principal components for dimension reduction, termed tangent PCA or TPCA, and study (i) Gaussian, (ii) mixture of Gaussian, and (iii) non-parametric densities to model the observed shapes. We validate these models using hypothesis testing, statistics of likelihood functions, and random sampling. It is demonstrated that a mixture of Gaussian model on TPCA captures best the observed shapes. Introduction Analysis of shapes is emerging as an important tool in recognition of objects from their images. As an example, one uses the contours formed by boundaries of objects, as they appear in images, to characterize the objects themselves. Since the objects can occur at arbitrary locations, scales, and planar rotations, without changing their appearances, one is interested in the shapes of these contours, rather than the contours themselves. This motivates the development of tools for statistical analysis of shapes of simple, closed curves in R.Astatistical analysis is beneficial in many situations. For instance, in cases where the observed image is low quality due to clutter, low resolution, or obscuration, one can use the contextual knowledge to impose prior models on expected shapes, and use a Bayesian framework to improve shape extraction performance. Such applications require a broad array of tools for analyzing shapes: geometric representations of shapes, metrics for quantifying shape differences, algorithms for computing shape statistics such as means and covariances, and tools for testing competing hypotheses on given shapes. Analysis of shapes of planar curves has been of a particular interest recently in the literature. Klassen [] have described a geometric technique to parameterize curves by their arc lengths, and to use their angle functions to represent P.J. Narayanan et al. (Eds.): ACCV, LNCS 8, pp.,. c Springer-Verlag Berlin Heidelberg

2 Statistical Shape Models Using Elastic-String Representations and analyze shapes. Similar constructions for analysis of closed curves were also studied in [, ]. Using the representations and metrics described in [], [] describe techniques for clustering, learning, and testing of planar shapes. One major limitation of this approach is that all curves are parameterized by arc length, and the resulting transformations from one shape into another are restricted to bending only. Local stretching or compression of shapes is not allowed. Mio [] resolved this issue by introducing a representation that allows both bending and stretching of shapes to match each other. The geodesic paths resulting from this approach seem more natural as interesting features, such as corners, are better preserved while constructing geodesics, in this approach. This representation of planar shapes is called an elastic string model. Our goal in this paper is to use elastic string model to study several probability models for capturing observed shape variability. Similar to approaches presented in [, ], we project observed shapes onto the tangent spaces at sample means, and further reduce their dimensions using PCA. Thus, we obtain a low-dimensional representations of shapes called TPCA. On tangent principal components (TPCs) of observed shapes we study: (i) Gaussian, (ii) nonparametric, and (iii) mixture of Gaussian models. The first two have been studied earlier for non-elastic shapes in []. To study model performances, we: (i) synthesize random shapes from these models, (ii) test amongst competing models using likelihood ratio, and (iii) compare statistics of likelihood on training and test data. This framework leads to stochastic shape models that can be used as priors in future Bayesian extraction of shapes from low-quality images. To illustrate these ideas we have used shapes from the ETH databases. Rest of this paper is organized as follows. Section summarizes elastic-string models for shape representations. Section proposes three candidate probability models for capturing shape variability, while Sections and study these probability models via synthesis and hypothesis testing. Elastic Strings Representation Here we summarize the main ideas behind elastic-string representations of planar shapes, originally described in Mio et al [].. Shape Representation Let α: [, π] R be a smooth parametric curve such that α (t), t [, π]. The velocity vector is α (t) = e φ(t) e jθ(t), where φ: [, π] R and θ :[, π] R are smooth, and j =. The function φ is the log-speed of α and θ is the angle function. φ(t) measures the rate at which the interval [, π] is stretched or compressed at t to form the curve α; φ(t) > indicates local stretching near t, andφ(t) < local compression. Curves parameterized by arc length have φ. We will represent α via the pair (φ, θ) and denote by H the collection of all such pairs. Parametric curves that differ by rigid motions or uniform scalings of the plane, or by re-parameterizations are treated as representing the same shape. The pair

3 A. Srivastava et al. (φ, θ) is already invariant to translations of the curve. Rigid rotations and uniform scalings are removed by restricting to the space, C = {(φ, θ) H : Z π e φ(t) dt =π, π Z π θ(t)e φ(t) dt = π, Z π e φ(t) e jθ(t) dt =}, C is called the pre-shape spaces of planar elastic strings. There are two possible ways of re-parameterizing a closed curve, without changing its shape: (i) One is to change the placement of origin t = on the curve. This change can be represented as the action of a unit circle S on a shape (φ, θ), according to: s (φ(t),θ(t)) = (φ(t s),θ(t s) +s). (ii) Re-parameterizations of α that preserve orientation and the property that α (t), t, are those obtained by composing α with an orientation-preserving diffeomorphism γ :[, π] [, π]. Let D be the group of all such mappings. These mappings define a right action of D on H by (φ, θ) γ =(φ γ +logγ,θ γ). () denotes composition of functions. The space of all (shape-preserving) reparametrization of a shape in C is thus given by S D. The resulting shape space is the space of all equivalence classes induced by these shape preserving transformations. It can be written as a quotient space S =(C/D)/S. What metric can used to compare shapes in this space? Mio [] suggests that, given (φ, θ) H, leth i and f i, i =,, represent infinitesimal deformations of φ and θ, resp., so that (h,f )and(h,f ) are tangent vectors to H at (φ, θ). For a, b >, define (h,f ), (h,f ) (φ,θ) as a h (t)h (t) e φ(t) dt + b f (t)f (t) e φ(t) dt. () It can be shown that re-parameterizations preserve the inner product, i.e., S D acts on H by isometries. The elastic properties of the curves are built-in to the model via the parameters a and b, which can be interpreted as tension and rigidity coefficients, respectively. Large values of the ratio a/b indicate that strings offer higher resistance to stretching and compression than to bending; the opposite holds for a/b small. In this paper we fix a value of a/b that balances between bending and stretching.. Geodesic Paths in Shape Spaces An important tool in this shape analysis is to construct geodesic paths, i.e. paths of smallest lengths, between arbitrary two shapes. Given the complicated geometry of S, this task is not straightforward, at least not analytically. One solution is to use a computational approach, where the search for geodesics is treated as an optimization problem with iterative numerical updates. This approach is called the shooting method.givenapairofshapesα (φ,θ )and α (φ,θ ), one solves: min Ψ (α ; g) (s (α )) γ () s S,γ D,g T α (C)

4 Statistical Shape Models Using Elastic-String Representations where Ψ t (α; g) denotes a geodesic path startingatashapeα in the direction g, and parameterized by time t. Also, is the L norm on H. Basically, one solves for the shooting direction g such that the geodesic from α in the direction g gets as close to the orbit of α under shape preserving transformations []. Let d(α,α ) g denote the length of geodesics connecting the shapes α and α. This construction helps define the exponential map: exp α (g) =Ψ (α; g) and its inverse exp α (β) =g such that Ψ (α; g) =β.. Sample Mean of Shapes Since the shape space S is nonlinear, the definitions of sample statistics, such as means and covariances, are not conventional. Earlier papers [7, 8] suggest the use of Karcher mean to define mean shapes as follows. For α,...,α n in S, and d(α i,α j ) the geodesic length between α i and α j, the Karcher mean is defined as the element µ Sthat minimizes the quantity n i= d(µ, α i). A gradientbased, iterative algorithm for computing the Karcher mean is presented in [8, ]. Shown in Figure are some examples of three classes of shapes dogs, pears, and mugs used in the experiments here, and the Figure shows Karcher means of shapes in these three classes. Let µ be the mean shape and for any shape α, let g T µ (S) be such that Ψ (µ; g) =α. Then, α called the exponential of g, i.e. exp µ (g), and conversely, g =exp µ (α). As described next, statistics of α are studied through statistics of its map onto the tangent space at the mean. Fig.. Examples of three classes of shapes dogs, pears, and mugs from the ETH database that are studied in this paper, with the numbers used in test and training Statistical Shape Models Our goal is to derive and analyze probability models for capturing observed shapes. The task of learning probability models on spaces like S is difficult for two main reasons. Firstly, they are nonlinear spaces and therefore classical statistical approaches, associated with the vector spaces, do not apply directly. Secondly, these are infinite-dimensional spaces and do not allow component-by-component modeling that is traditionally followed in finite-dimensional vector spaces. The solution involves making two approximations. First, we project elements of S onto the tangent space T µ (S), which is a vector space, and therefore, better suited to statistical modeling. This is performed using the inverse exponential map exp µ. Second, we perform dimension reduction in T µ (S) using PCA. Together,

5 A. Srivastava et al. Fig.. In each case, left image shows the Karcher mean of shapes and right shows plots of the singular values of sample covariance matrix these two approximations given rise to TPCA representation. These ideas were first proposed for landmark-based shape analysis in []. To start TPCA, we use the Gram-Schmidt algorithm to find an orthonormal basis of the given vectors: Set i =andr =.. Set Y i = g r i j= Y j,g r Y j.. If Y i,y i, Set Y i = Y i / Y i,y i, i = i +,r = r +, and go to Step. Else If r<k Set r = r +andgotostep. Else Stop Say the algorithm stops at some i = n k. Sonowwehaveann-dimensional subspace Y spanned by an orthonormal basis with elements {Y,Y,...,Y n }.The next step is to project each of the observed vector into Y as follows. Let x ij = g i,y j and define a vector x i =[x i,x i,...,x in ] R n. Then, the projection of g i into Y is given by n j= x ijy j.eachg i T µ (S) is now represented by a smaller vector x i R n.next,weperformpcainr n using the projected observations {x, x,...,x k }. That is, from their sample covariance matrix C R n n, find its singular value decomposition C = UΣU T, and use the first d-columns of U to form a basis for the principal subspace of R n,withd n. The vector x R n maps to a smaller vector a R d such that x = d j= a ju j. The choice of d is made using the singular values of C; shown in Figure are the plots of singular values of C for the three classes: dogs, pears, and mugs.. Probability Models on TPCs We impose a probability model on α implicitly by imposing a probability model on its tangent principal components (TPCs) a. What probability models can be used in this situation? In this paper, we study the following three models: nonparametric, Gaussian and mixtures of Gaussian. The first two models were studied for non-elastic shapes in [].. Nonparametric Model: Assuming that the TPCs, a j s, are statistically independent of each other, one can estimate their probability densities directly from the data using a kernel estimator. Let f () j, j =,...,d be the kernel estimate of the density function of a j,thej th TPC of the shape α. Intheexperiments presented here we used a Gaussian Kernel. Then, assuming independence of

6 Statistical Shape Models Using Elastic-String Representations 7 TPCs, we obtain: f () (α) = d j= f j(a j ). Shown in Figure are some examples of estimated f () for several js. For each shape class, we display three examples of non-parametric density estimates for modeling TPCs. Dogs Pears Mugs Fig.. We show three examples of modeling TPCs in each class. For each example, the left figure shows nonparametric estimate f () while the right figure shows the mixture of Gaussian f () (using cross-lines) drawn over observed densities (plain lines).. Gaussian Model: Let Σ R d d be the diagonal matrix in SVD of C, the sample covariance of x i s. Then, we can model the component a j as a Gaussian random variable with mean zero and variance Σ jj. Denoting the Gaussian density function as h(y; z,σ ) exp( (y πσ z) /(σ )), we obtain the Gaussian shape model f () (α) = d j= h(a j;,σ jj ).. Mixture of Gaussian: Another candidate model is that a j follows the density ( d K ) f () j (α) = p k h(a j ; z k,σk), p k =, j= k= a finite mixture of Gaussian. For a given K, EMalgorithmcanbeusedto estimate the means and variances of components. Based on empirical evidence, we have used K = in this paper to estimate f () from observed data. Shown in Figure are some examples of estimated f () for some TPCs. In each panel, the marked line shows the estimated mixture density and, for comparison, the plane line shows the observed histograms. k Empirical Evaluations We have analyzed and validated the proposed shape models using: (i) random sampling, (ii) hypothesis testing, and (iii) statistics of log-likelihoods. We describe these results next.

7 8 A. Srivastava et al. Fig.. Sample shapes synthesized from the nonparametric model (top) and the mixture model (bottom) Shape Sampling: As a first step, we have synthesized random shapes from the three probability models f (i), i =,,. In each case the synthesis involves generating a random TPC according to its probability model- kernel density, Gaussian density or mixture of Gaussian- and then reconstructing the shape represented by that set of TPCs. For the generated values of TPCs, we form the vector x = d j= a ju j, and the tangent direction g = n i= x iy i, and eventually the shape α = exp µ (g). Shown in Figure are examples of random shapes generated from the models f () (top row) and f () (bottom row). We found that all three models seem to perform reasonably well in synthesis, with f () and f () being slightly better than f (). Testing Shape Models In order to test proposed models for capturing observed shape variability, we use the likelihood ratio test to select among the candidate models. For a shape α S, the likelihood ratio under any two models is: f (m) (α) f (n) (α) = and the log-likelihood ratio is l(α; m, n) d f (m) j (a j ), m,n =,,, f (n) j (a j ) j= d ( ) log(f (m) j (a j )) log(f (n) j (a j )). j= If l(α; m, n) is positive then the model m is selected, and vice-versa. Taking a large set of test shapes, we have evaluated l(α; m, n) for each shape and have counted the fraction for which l(α; m, n) is positive. We define: P (m, n) = {i l(α i; m, n) > } k where k is the total number of shapes used in this test. This fraction is plotted versus the component size d in Figure, for two pairs of shape models: P (, ) in,

8 Statistical Shape Models Using Elastic-String Representations dogs pears mugs Fig.. P (m, n) plotted versus vs d, for each of the three classes. Top row: m =, n =, and bottom row: m =,n =. the top row and P (, ) in the bottom row. P (m, n) >. implies that model m outperforms n. Two sets of results are presented in each of these plots. The solid line is for the test shapes that were not used in estimation of shape models, and the broken line is for the training shapes that were used in model estimation. Also, we draw a line at. to clarify which model is performing better. As these indicate, the mixture model seems to perform the best in most situations. On the training shapes, for pears and mugs, the nonparametric model is better than the mixture model. This result in expected since nonparametric model is derived from these training shapes themselves. However, on the test shapes, the mixture model is either comparable or better than the other two models. We conclude that for this data set, the mixture model is better for capturing variability in both training and test shapes. Furthermore, it is efficient due to its parametric nature. Statistics of Model Likelihoods Another technique for model validation is to study the variability of a modelbased sufficient statistic when evaluatedonbothtrainingandtestshapes.in case the distributions of this statistic are similar on both training and test shapes, this validates the underlying model. In this paper, we have chosen the sufficient dogs pears mugs dogs pears mugs Fig.. Histograms of ν (i) (α) for test (solid) and training shapes (broken). First three are for nonparametric model, and the last three are for mixture of Gaussians.

9 A. Srivastava et al. statistictobeproportionaltonegative log-likelihood of an observed shape. That is, we define ν (i) (α) log(f (i) (α)), where the proportionality implies that the constants have been ignored. Shown in Figure are some examples of this study for the nonparametric (first three) and the mixture model (last three). These plots shows histograms of ν (i) (α) values for both test and training shapes, for each of the three shape classes. It is evident that the histograms for training and test sets are quite similar in all these examples, and hence, validate the proposed models. Acceptance/Rejection Under Learned Models: In the final experiment, we performed acceptance/rejection for each test shape under the mixture model, f (), for each shape class (dogs, pears, and mugs). Using threshold values estimated using training data of each class, we compute the value of ν () (α) foreach test shape α; if it is below the threshold we accept it, otherwise we reject it. For example, we have dog reject ν () > (α) < κ dog. dog accept This is done for each of the three classes dogs, pears, and mugs, and the results are summarized in the next table. This table lists the percentage of times a shape from a given test class was accepted by each of the three shape classes. For example, test shapes in dog class were accepted 9.7% times by shape model for dog class,.7% by pear model, and.8% by cup model. Also,.7% of test shapes in dog class were rejected by all three models. Since a shape can be accepted by more than one model, the sum in each row can exceed %. Notice that the test shapes also include other objects such as horses, cows, apples, cars, and tomatoes. Some of the cows (%) are accepted under dog model, but are easily rejected under pear and mug models; most of the cows (%) are rejected under all three models. Tomatoes are mostly accepted by pear and mug models. Overall, the mixture model f () demonstrates a significant success in capturing shape variability and in discriminating between object classes. It also enjoys the efficiency of being a parametric model. Test class Dog Accepts (%) Pear Accepts (%) Cups Accepts (%) No Accepts (%) Dogs Pears Cups Horses Apples Cows Cars Tomatoes

10 Conclusion Statistical Shape Models Using Elastic-String Representations We have presented results from statistical analysis of planar shapes under elastic string models. Using TPCA representation of shapes, three candidate models were presented: nonparametric, Gaussian, and a mixture of Gaussian. We evaluated these models using (i) random sampling, (ii) likelihood ratio tests, (iii) similarity of (distributions of) sufficient statistics on training and test shapes, and (iv) acceptance/rejection of test shapes under the models estimated from the corresponding training shapes. All three models do reasonably well in random sampling and likelihood ratio test. However, the mixture model emerges as the best model for capturing shape variability and efficiency. We therefore conjecture that mixture of Gaussians are sufficient for modeling TPCs of observed shapes for use as prior shape models in future Bayesian inferences. References. Klassen, E., Srivastava, A., Mio, W., Joshi, S.: Analysis of planar shapes using geodesic paths on shape spaces. IEEE Pattern Analysis and Machine Intelligence (March, ) 7 8. Younes, L.: Optimal matching between shapes via elastic deformations. Journal of Image and Vision Computing 7 (999) Michor, P.W., Mumford, D.: Riemannian geometries on spaces of plane curves. Journal of the European Mathematical Society to appear (). Srivastava, A., Joshi, S., Mio, W., Liu, X.: Statistical shape analysis: Clustering, learning and testing. IEEE Transactions on Pattern Analysis and Machine Intelligence 7 () 9. Mio, W., Srivastava, A.: Elastic string models for representation and analysis of planar shapes. In: Proc. of IEEE Computer Vision and Pattern Recognition. (). Dryden, I.L., Mardia, K.V.: Statistical Shape Analysis. John Wiley & Son (998) 7. Le, H.L., Kendall, D.G.: The Riemannian structure of Euclidean shape spaces: a novel environment for statistics. Annals of Statistics (99) 7 8. Karcher, H.: Riemann center of mass and mollifier smoothing. Communications on PureandAppliedMathematics (977) 9

Statistical Shape Analysis: Clustering, Learning, and Testing

Statistical Shape Analysis: Clustering, Learning, and Testing Statistical Shape Analysis: Clustering, Learning, and Testing Anuj Srivastava Shantanu Joshi Washington Mio Xiuwen Liu Abstract Using a differential-geometric treatment of planar shapes, we present tools

More information

A GENTLE INTRODUCTION TO THE BASIC CONCEPTS OF SHAPE SPACE AND SHAPE STATISTICS

A GENTLE INTRODUCTION TO THE BASIC CONCEPTS OF SHAPE SPACE AND SHAPE STATISTICS A GENTLE INTRODUCTION TO THE BASIC CONCEPTS OF SHAPE SPACE AND SHAPE STATISTICS HEMANT D. TAGARE. Introduction. Shape is a prominent visual feature in many images. Unfortunately, the mathematical theory

More information

TOOLS FOR ANALYZING SHAPES OF CURVES AND SURFACES. ANUJ SRIVASTAVA Department of Statistics Florida State University

TOOLS FOR ANALYZING SHAPES OF CURVES AND SURFACES. ANUJ SRIVASTAVA Department of Statistics Florida State University TOOLS FOR ANALYZING SHAPES OF CURVES AND SURFACES ANUJ SRIVASTAVA Department of Statistics Florida State University 1 MURI: INTGERATED FUSION AND SENSOR MANAGEMENT FOR ATE Innovative Front-End Processing

More information

Aggregated Pairwise Classification of Statistical Shapes

Aggregated Pairwise Classification of Statistical Shapes Aggregated Pairwise Classification of Statistical Shapes arxiv:1901.07593v1 [stat.ml] 22 Jan 2019 Min Ho Cho, Sebastian Kurtek, and Steven N. MacEachern Department of Statistics, The Ohio State University

More information

A Computational Approach to Fisher Information Geometry with Applications to Image Analysis

A Computational Approach to Fisher Information Geometry with Applications to Image Analysis A Computational Approach to Fisher Information Geometry with Applications to Image Analysis Washington Mio, Dennis Badlyans, and Xiuwen Liu Department of Mathematics, Florida State University Tallahassee,

More information

Unsupervised Learning

Unsupervised Learning Unsupervised Learning Learning without Class Labels (or correct outputs) Density Estimation Learn P(X) given training data for X Clustering Partition data into clusters Dimensionality Reduction Discover

More information

Unsupervised learning in Vision

Unsupervised learning in Vision Chapter 7 Unsupervised learning in Vision The fields of Computer Vision and Machine Learning complement each other in a very natural way: the aim of the former is to extract useful information from visual

More information

Geodesic and parallel models for leaf shape

Geodesic and parallel models for leaf shape Geodesic and parallel models for leaf shape Stephan F. Huckemann and Thomas Hotz Institute for Mathematical Stochastics, Georg-August Universität Göttingen 1 Introduction Since more than a decade, many

More information

Landmark-Constrained Elastic Shape Analysis of Planar Curves

Landmark-Constrained Elastic Shape Analysis of Planar Curves Landmark-Constrained Elastic Shape Analysis of Planar Curves Justin Strait, Sebastian Kurtek, Emily Bartha, Steven MacEachern Department of Statistics, The Ohio State University April 6, 2017 Abstract

More information

Automatic 3D Face Recognition Using Shapes of Facial Curves

Automatic 3D Face Recognition Using Shapes of Facial Curves Automatic 3D Face Recognition Using Shapes of Facial Curves Chafik Samir, Anuj Srivastava, and Mohamed Daoudi Abstract In addition to facial textures, manifested in facial images, shapes of facial surfaces

More information

MULTIVARIATE TEXTURE DISCRIMINATION USING A PRINCIPAL GEODESIC CLASSIFIER

MULTIVARIATE TEXTURE DISCRIMINATION USING A PRINCIPAL GEODESIC CLASSIFIER MULTIVARIATE TEXTURE DISCRIMINATION USING A PRINCIPAL GEODESIC CLASSIFIER A.Shabbir 1, 2 and G.Verdoolaege 1, 3 1 Department of Applied Physics, Ghent University, B-9000 Ghent, Belgium 2 Max Planck Institute

More information

Globally Stabilized 3L Curve Fitting

Globally Stabilized 3L Curve Fitting Globally Stabilized 3L Curve Fitting Turker Sahin and Mustafa Unel Department of Computer Engineering, Gebze Institute of Technology Cayirova Campus 44 Gebze/Kocaeli Turkey {htsahin,munel}@bilmuh.gyte.edu.tr

More information

Shape analysis through geometric distributions

Shape analysis through geometric distributions Shape analysis through geometric distributions Nicolas Charon (CIS, Johns Hopkins University) joint work with B. Charlier (Université de Montpellier), I. Kaltenmark (CMLA), A. Trouvé, Hsi-Wei Hsieh (JHU)...

More information

Dimension Reduction CS534

Dimension Reduction CS534 Dimension Reduction CS534 Why dimension reduction? High dimensionality large number of features E.g., documents represented by thousands of words, millions of bigrams Images represented by thousands of

More information

Intrinsic Mean for Semi-metrical Shape Retrieval Via Graph Cuts

Intrinsic Mean for Semi-metrical Shape Retrieval Via Graph Cuts Intrinsic Mean for Semi-metrical Shape Retrieval Via Graph Cuts Frank R. Schmidt 1,EnoTöppe 1,DanielCremers 1,andYuriBoykov 2 1 Department of Computer Science University of Bonn Römerstr. 164, 53117 Bonn,

More information

Generative and discriminative classification techniques

Generative and discriminative classification techniques Generative and discriminative classification techniques Machine Learning and Category Representation 013-014 Jakob Verbeek, December 13+0, 013 Course website: http://lear.inrialpes.fr/~verbeek/mlcr.13.14

More information

Spline Curves. Spline Curves. Prof. Dr. Hans Hagen Algorithmic Geometry WS 2013/2014 1

Spline Curves. Spline Curves. Prof. Dr. Hans Hagen Algorithmic Geometry WS 2013/2014 1 Spline Curves Prof. Dr. Hans Hagen Algorithmic Geometry WS 2013/2014 1 Problem: In the previous chapter, we have seen that interpolating polynomials, especially those of high degree, tend to produce strong

More information

AN important goal in image analysis is to classify and

AN important goal in image analysis is to classify and 590 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 27, NO. 4, APRIL 2005 Statistical Shape Analysis: Clustering, Learning, and Testing Anuj Srivastava, Member, IEEE, Shantanu H. Joshi,

More information

Introduction to geometry

Introduction to geometry 1 2 Manifolds A topological space in which every point has a neighborhood homeomorphic to (topological disc) is called an n-dimensional (or n-) manifold Introduction to geometry The German way 2-manifold

More information

7.1 INTRODUCTION Wavelet Transform is a popular multiresolution analysis tool in image processing and

7.1 INTRODUCTION Wavelet Transform is a popular multiresolution analysis tool in image processing and Chapter 7 FACE RECOGNITION USING CURVELET 7.1 INTRODUCTION Wavelet Transform is a popular multiresolution analysis tool in image processing and computer vision, because of its ability to capture localized

More information

Bayesian Active Contours with Affine-Invariant, Elastic Shape Prior

Bayesian Active Contours with Affine-Invariant, Elastic Shape Prior Bayesian Active Contours with Affine-Invariant, Elastic Shape Prior Darshan Bryner and Anuj Srivastava Department of Statistics Florida State University, Tallahassee, FL {dbryner,anuj}@stat.fsu.edu Abstract

More information

04 - Normal Estimation, Curves

04 - Normal Estimation, Curves 04 - Normal Estimation, Curves Acknowledgements: Olga Sorkine-Hornung Normal Estimation Implicit Surface Reconstruction Implicit function from point clouds Need consistently oriented normals < 0 0 > 0

More information

Curve Subdivision in SE(2)

Curve Subdivision in SE(2) Curve Subdivision in SE(2) Jan Hakenberg, ETH Zürich 2018-07-26 Figure: A point in the special Euclidean group SE(2) consists of a position in the plane and a heading. The figure shows two rounds of cubic

More information

Convexization in Markov Chain Monte Carlo

Convexization in Markov Chain Monte Carlo in Markov Chain Monte Carlo 1 IBM T. J. Watson Yorktown Heights, NY 2 Department of Aerospace Engineering Technion, Israel August 23, 2011 Problem Statement MCMC processes in general are governed by non

More information

Fingerprint Classification Using Orientation Field Flow Curves

Fingerprint Classification Using Orientation Field Flow Curves Fingerprint Classification Using Orientation Field Flow Curves Sarat C. Dass Michigan State University sdass@msu.edu Anil K. Jain Michigan State University ain@msu.edu Abstract Manual fingerprint classification

More information

DRAFT: Analysis of Skew Tensegrity Prisms

DRAFT: Analysis of Skew Tensegrity Prisms DRAFT: Analysis of Skew Tensegrity Prisms Mark Schenk March 6, 2006 Abstract This paper describes the properties of skew tensegrity prisms. By showing that the analytical equilibrium solutions of regular

More information

Dimension Reduction of Image Manifolds

Dimension Reduction of Image Manifolds Dimension Reduction of Image Manifolds Arian Maleki Department of Electrical Engineering Stanford University Stanford, CA, 9435, USA E-mail: arianm@stanford.edu I. INTRODUCTION Dimension reduction of datasets

More information

Shape-based Diffeomorphic Registration on Hippocampal Surfaces Using Beltrami Holomorphic Flow

Shape-based Diffeomorphic Registration on Hippocampal Surfaces Using Beltrami Holomorphic Flow Shape-based Diffeomorphic Registration on Hippocampal Surfaces Using Beltrami Holomorphic Flow Abstract. Finding meaningful 1-1 correspondences between hippocampal (HP) surfaces is an important but difficult

More information

DETC APPROXIMATE MOTION SYNTHESIS OF SPHERICAL KINEMATIC CHAINS

DETC APPROXIMATE MOTION SYNTHESIS OF SPHERICAL KINEMATIC CHAINS Proceedings of the ASME 2007 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference IDETC/CIE 2007 September 4-7, 2007, Las Vegas, Nevada, USA DETC2007-34372

More information

STATISTICS AND ANALYSIS OF SHAPE

STATISTICS AND ANALYSIS OF SHAPE Control and Cybernetics vol. 36 (2007) No. 2 Book review: STATISTICS AND ANALYSIS OF SHAPE by H. Krim, A. Yezzi, Jr., eds. There are numerous definitions of a notion of shape of an object. These definitions

More information

Non-linear dimension reduction

Non-linear dimension reduction Sta306b May 23, 2011 Dimension Reduction: 1 Non-linear dimension reduction ISOMAP: Tenenbaum, de Silva & Langford (2000) Local linear embedding: Roweis & Saul (2000) Local MDS: Chen (2006) all three methods

More information

Planes Intersecting Cones: Static Hypertext Version

Planes Intersecting Cones: Static Hypertext Version Page 1 of 12 Planes Intersecting Cones: Static Hypertext Version On this page, we develop some of the details of the plane-slicing-cone picture discussed in the introduction. The relationship between the

More information

Level-set MCMC Curve Sampling and Geometric Conditional Simulation

Level-set MCMC Curve Sampling and Geometric Conditional Simulation Level-set MCMC Curve Sampling and Geometric Conditional Simulation Ayres Fan John W. Fisher III Alan S. Willsky February 16, 2007 Outline 1. Overview 2. Curve evolution 3. Markov chain Monte Carlo 4. Curve

More information

Estimating normal vectors and curvatures by centroid weights

Estimating normal vectors and curvatures by centroid weights Computer Aided Geometric Design 21 (2004) 447 458 www.elsevier.com/locate/cagd Estimating normal vectors and curvatures by centroid weights Sheng-Gwo Chen, Jyh-Yang Wu Department of Mathematics, National

More information

Last week. Multi-Frame Structure from Motion: Multi-View Stereo. Unknown camera viewpoints

Last week. Multi-Frame Structure from Motion: Multi-View Stereo. Unknown camera viewpoints Last week Multi-Frame Structure from Motion: Multi-View Stereo Unknown camera viewpoints Last week PCA Today Recognition Today Recognition Recognition problems What is it? Object detection Who is it? Recognizing

More information

Image Enhancement Techniques for Fingerprint Identification

Image Enhancement Techniques for Fingerprint Identification March 2013 1 Image Enhancement Techniques for Fingerprint Identification Pankaj Deshmukh, Siraj Pathan, Riyaz Pathan Abstract The aim of this paper is to propose a new method in fingerprint enhancement

More information

Shape Classification and Cell Movement in 3D Matrix Tutorial (Part I)

Shape Classification and Cell Movement in 3D Matrix Tutorial (Part I) Shape Classification and Cell Movement in 3D Matrix Tutorial (Part I) Fred Park UCI icamp 2011 Outline 1. Motivation and Shape Definition 2. Shape Descriptors 3. Classification 4. Applications: Shape Matching,

More information

( ) =cov X Y = W PRINCIPAL COMPONENT ANALYSIS. Eigenvectors of the covariance matrix are the principal components

( ) =cov X Y = W PRINCIPAL COMPONENT ANALYSIS. Eigenvectors of the covariance matrix are the principal components Review Lecture 14 ! PRINCIPAL COMPONENT ANALYSIS Eigenvectors of the covariance matrix are the principal components 1. =cov X Top K principal components are the eigenvectors with K largest eigenvalues

More information

Bayesian Spherical Wavelet Shrinkage: Applications to Shape Analysis

Bayesian Spherical Wavelet Shrinkage: Applications to Shape Analysis Bayesian Spherical Wavelet Shrinkage: Applications to Shape Analysis Xavier Le Faucheur a, Brani Vidakovic b and Allen Tannenbaum a a School of Electrical and Computer Engineering, b Department of Biomedical

More information

Topological Mapping. Discrete Bayes Filter

Topological Mapping. Discrete Bayes Filter Topological Mapping Discrete Bayes Filter Vision Based Localization Given a image(s) acquired by moving camera determine the robot s location and pose? Towards localization without odometry What can be

More information

The Role of Manifold Learning in Human Motion Analysis

The Role of Manifold Learning in Human Motion Analysis The Role of Manifold Learning in Human Motion Analysis Ahmed Elgammal and Chan Su Lee Department of Computer Science, Rutgers University, Piscataway, NJ, USA {elgammal,chansu}@cs.rutgers.edu Abstract.

More information

COMPUTER AND ROBOT VISION

COMPUTER AND ROBOT VISION VOLUME COMPUTER AND ROBOT VISION Robert M. Haralick University of Washington Linda G. Shapiro University of Washington T V ADDISON-WESLEY PUBLISHING COMPANY Reading, Massachusetts Menlo Park, California

More information

1. Carlos A. Felippa, Introduction to Finite Element Methods,

1. Carlos A. Felippa, Introduction to Finite Element Methods, Chapter Finite Element Methods In this chapter we will consider how one can model the deformation of solid objects under the influence of external (and possibly internal) forces. As we shall see, the coupled

More information

Probabilistic Facial Feature Extraction Using Joint Distribution of Location and Texture Information

Probabilistic Facial Feature Extraction Using Joint Distribution of Location and Texture Information Probabilistic Facial Feature Extraction Using Joint Distribution of Location and Texture Information Mustafa Berkay Yilmaz, Hakan Erdogan, Mustafa Unel Sabanci University, Faculty of Engineering and Natural

More information

An Efficient Model Selection for Gaussian Mixture Model in a Bayesian Framework

An Efficient Model Selection for Gaussian Mixture Model in a Bayesian Framework IEEE SIGNAL PROCESSING LETTERS, VOL. XX, NO. XX, XXX 23 An Efficient Model Selection for Gaussian Mixture Model in a Bayesian Framework Ji Won Yoon arxiv:37.99v [cs.lg] 3 Jul 23 Abstract In order to cluster

More information

CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS

CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS CHAPTER 4 CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS 4.1 Introduction Optical character recognition is one of

More information

GLOBAL SAMPLING OF IMAGE EDGES. Demetrios P. Gerogiannis, Christophoros Nikou, Aristidis Likas

GLOBAL SAMPLING OF IMAGE EDGES. Demetrios P. Gerogiannis, Christophoros Nikou, Aristidis Likas GLOBAL SAMPLING OF IMAGE EDGES Demetrios P. Gerogiannis, Christophoros Nikou, Aristidis Likas Department of Computer Science and Engineering, University of Ioannina, 45110 Ioannina, Greece {dgerogia,cnikou,arly}@cs.uoi.gr

More information

Chapter 11 Representation & Description

Chapter 11 Representation & Description Chain Codes Chain codes are used to represent a boundary by a connected sequence of straight-line segments of specified length and direction. The direction of each segment is coded by using a numbering

More information

Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial Region Segmentation

Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial Region Segmentation IJCSNS International Journal of Computer Science and Network Security, VOL.13 No.11, November 2013 1 Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial

More information

Linear Methods for Regression and Shrinkage Methods

Linear Methods for Regression and Shrinkage Methods Linear Methods for Regression and Shrinkage Methods Reference: The Elements of Statistical Learning, by T. Hastie, R. Tibshirani, J. Friedman, Springer 1 Linear Regression Models Least Squares Input vectors

More information

Range Image Registration with Edge Detection in Spherical Coordinates

Range Image Registration with Edge Detection in Spherical Coordinates Range Image Registration with Edge Detection in Spherical Coordinates Olcay Sertel 1 and Cem Ünsalan2 Computer Vision Research Laboratory 1 Department of Computer Engineering 2 Department of Electrical

More information

IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS /$ IEEE

IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS /$ IEEE IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS 1 Exploration of Heterogeneous FPGAs for Mapping Linear Projection Designs Christos-S. Bouganis, Member, IEEE, Iosifina Pournara, and Peter

More information

Planar Symmetry Detection by Random Sampling and Voting Process

Planar Symmetry Detection by Random Sampling and Voting Process Planar Symmetry Detection by Random Sampling and Voting Process Atsushi Imiya, Tomoki Ueno, and Iris Fermin Dept. of IIS, Chiba University, 1-33, Yayo-cho, Inage-ku, Chiba, 263-8522, Japan imiya@ics.tj.chiba-u.ac.jp

More information

Epitomic Analysis of Human Motion

Epitomic Analysis of Human Motion Epitomic Analysis of Human Motion Wooyoung Kim James M. Rehg Department of Computer Science Georgia Institute of Technology Atlanta, GA 30332 {wooyoung, rehg}@cc.gatech.edu Abstract Epitomic analysis is

More information

An Adaptive Eigenshape Model

An Adaptive Eigenshape Model An Adaptive Eigenshape Model Adam Baumberg and David Hogg School of Computer Studies University of Leeds, Leeds LS2 9JT, U.K. amb@scs.leeds.ac.uk Abstract There has been a great deal of recent interest

More information

Region-based Segmentation

Region-based Segmentation Region-based Segmentation Image Segmentation Group similar components (such as, pixels in an image, image frames in a video) to obtain a compact representation. Applications: Finding tumors, veins, etc.

More information

Texture Mapping using Surface Flattening via Multi-Dimensional Scaling

Texture Mapping using Surface Flattening via Multi-Dimensional Scaling Texture Mapping using Surface Flattening via Multi-Dimensional Scaling Gil Zigelman Ron Kimmel Department of Computer Science, Technion, Haifa 32000, Israel and Nahum Kiryati Department of Electrical Engineering

More information

Shape Context Matching For Efficient OCR

Shape Context Matching For Efficient OCR Matching For Efficient OCR May 14, 2012 Matching For Efficient OCR Table of contents 1 Motivation Background 2 What is a? Matching s Simliarity Measure 3 Matching s via Pyramid Matching Matching For Efficient

More information

Note Set 4: Finite Mixture Models and the EM Algorithm

Note Set 4: Finite Mixture Models and the EM Algorithm Note Set 4: Finite Mixture Models and the EM Algorithm Padhraic Smyth, Department of Computer Science University of California, Irvine Finite Mixture Models A finite mixture model with K components, for

More information

Clustering K-means. Machine Learning CSEP546 Carlos Guestrin University of Washington February 18, Carlos Guestrin

Clustering K-means. Machine Learning CSEP546 Carlos Guestrin University of Washington February 18, Carlos Guestrin Clustering K-means Machine Learning CSEP546 Carlos Guestrin University of Washington February 18, 2014 Carlos Guestrin 2005-2014 1 Clustering images Set of Images [Goldberger et al.] Carlos Guestrin 2005-2014

More information

Structure from Motion. Prof. Marco Marcon

Structure from Motion. Prof. Marco Marcon Structure from Motion Prof. Marco Marcon Summing-up 2 Stereo is the most powerful clue for determining the structure of a scene Another important clue is the relative motion between the scene and (mono)

More information

Exponential Maps for Computer Vision

Exponential Maps for Computer Vision Exponential Maps for Computer Vision Nick Birnie School of Informatics University of Edinburgh 1 Introduction In computer vision, the exponential map is the natural generalisation of the ordinary exponential

More information

Background for Surface Integration

Background for Surface Integration Background for urface Integration 1 urface Integrals We have seen in previous work how to define and compute line integrals in R 2. You should remember the basic surface integrals that we will need to

More information

Robust Kernel Methods in Clustering and Dimensionality Reduction Problems

Robust Kernel Methods in Clustering and Dimensionality Reduction Problems Robust Kernel Methods in Clustering and Dimensionality Reduction Problems Jian Guo, Debadyuti Roy, Jing Wang University of Michigan, Department of Statistics Introduction In this report we propose robust

More information

The Curse of Dimensionality

The Curse of Dimensionality The Curse of Dimensionality ACAS 2002 p1/66 Curse of Dimensionality The basic idea of the curse of dimensionality is that high dimensional data is difficult to work with for several reasons: Adding more

More information

Application of Characteristic Function Method in Target Detection

Application of Characteristic Function Method in Target Detection Application of Characteristic Function Method in Target Detection Mohammad H Marhaban and Josef Kittler Centre for Vision, Speech and Signal Processing University of Surrey Surrey, GU2 7XH, UK eep5mm@ee.surrey.ac.uk

More information

Preliminaries: Size Measures and Shape Coordinates

Preliminaries: Size Measures and Shape Coordinates 2 Preliminaries: Size Measures and Shape Coordinates 2.1 Configuration Space Definition 2.1 The configuration is the set of landmarks on a particular object. The configuration matrix X is the k m matrix

More information

Model-based segmentation and recognition from range data

Model-based segmentation and recognition from range data Model-based segmentation and recognition from range data Jan Boehm Institute for Photogrammetry Universität Stuttgart Germany Keywords: range image, segmentation, object recognition, CAD ABSTRACT This

More information

Open-Curve Shape Correspondence Without Endpoint Correspondence

Open-Curve Shape Correspondence Without Endpoint Correspondence Open-Curve Shape Correspondence Without Endpoint Correspondence Theodor Richardson and Song Wang Department of Computer Science and Engineering, University of South Carolina, Columbia, SC 29208, USA richa268@cse.sc.edu,

More information

Motivation. Parametric Curves (later Surfaces) Outline. Tangents, Normals, Binormals. Arclength. Advanced Computer Graphics (Fall 2010)

Motivation. Parametric Curves (later Surfaces) Outline. Tangents, Normals, Binormals. Arclength. Advanced Computer Graphics (Fall 2010) Advanced Computer Graphics (Fall 2010) CS 283, Lecture 19: Basic Geometric Concepts and Rotations Ravi Ramamoorthi http://inst.eecs.berkeley.edu/~cs283/fa10 Motivation Moving from rendering to simulation,

More information

Florida State University Libraries

Florida State University Libraries Florida State University Libraries Electronic Theses, Treatises and Dissertations The Graduate School 213 Statistical Analysis of Trajectories on Riemannian Manifolds Jingyong Su Follow this and additional

More information

CONFORMAL MAPPING POTENTIAL FLOW AROUND A WINGSECTION USED AS A TEST CASE FOR THE INVISCID PART OF RANS SOLVERS

CONFORMAL MAPPING POTENTIAL FLOW AROUND A WINGSECTION USED AS A TEST CASE FOR THE INVISCID PART OF RANS SOLVERS European Conference on Computational Fluid Dynamics ECCOMAS CFD 2006 P. Wesseling, E. Oñate and J. Périaux (Eds) c TU Delft, The Netherlands, 2006 CONFORMAL MAPPING POTENTIAL FLOW AROUND A WINGSECTION

More information

Occluded Facial Expression Tracking

Occluded Facial Expression Tracking Occluded Facial Expression Tracking Hugo Mercier 1, Julien Peyras 2, and Patrice Dalle 1 1 Institut de Recherche en Informatique de Toulouse 118, route de Narbonne, F-31062 Toulouse Cedex 9 2 Dipartimento

More information

Locally Weighted Least Squares Regression for Image Denoising, Reconstruction and Up-sampling

Locally Weighted Least Squares Regression for Image Denoising, Reconstruction and Up-sampling Locally Weighted Least Squares Regression for Image Denoising, Reconstruction and Up-sampling Moritz Baecher May 15, 29 1 Introduction Edge-preserving smoothing and super-resolution are classic and important

More information

22 October, 2012 MVA ENS Cachan. Lecture 5: Introduction to generative models Iasonas Kokkinos

22 October, 2012 MVA ENS Cachan. Lecture 5: Introduction to generative models Iasonas Kokkinos Machine Learning for Computer Vision 1 22 October, 2012 MVA ENS Cachan Lecture 5: Introduction to generative models Iasonas Kokkinos Iasonas.kokkinos@ecp.fr Center for Visual Computing Ecole Centrale Paris

More information

Orientation of manifolds - definition*

Orientation of manifolds - definition* Bulletin of the Manifold Atlas - definition (2013) Orientation of manifolds - definition* MATTHIAS KRECK 1. Zero dimensional manifolds For zero dimensional manifolds an orientation is a map from the manifold

More information

The Pre-Image Problem in Kernel Methods

The Pre-Image Problem in Kernel Methods The Pre-Image Problem in Kernel Methods James Kwok Ivor Tsang Department of Computer Science Hong Kong University of Science and Technology Hong Kong The Pre-Image Problem in Kernel Methods ICML-2003 1

More information

Statistical Shape Analysis

Statistical Shape Analysis Statistical Shape Analysis I. L. Dryden and K. V. Mardia University ofleeds, UK JOHN WILEY& SONS Chichester New York Weinheim Brisbane Singapore Toronto Contents Preface Acknowledgements xv xix 1 Introduction

More information

Biometrics Technology: Image Processing & Pattern Recognition (by Dr. Dickson Tong)

Biometrics Technology: Image Processing & Pattern Recognition (by Dr. Dickson Tong) Biometrics Technology: Image Processing & Pattern Recognition (by Dr. Dickson Tong) References: [1] http://homepages.inf.ed.ac.uk/rbf/hipr2/index.htm [2] http://www.cs.wisc.edu/~dyer/cs540/notes/vision.html

More information

A Simple Method of the TEX Surface Drawing Suitable for Teaching Materials with the Aid of CAS

A Simple Method of the TEX Surface Drawing Suitable for Teaching Materials with the Aid of CAS A Simple Method of the TEX Surface Drawing Suitable for Teaching Materials with the Aid of CAS Masataka Kaneko, Hajime Izumi, Kiyoshi Kitahara 1, Takayuki Abe, Kenji Fukazawa 2, Masayoshi Sekiguchi, Yuuki

More information

CHAPTER 6 PERCEPTUAL ORGANIZATION BASED ON TEMPORAL DYNAMICS

CHAPTER 6 PERCEPTUAL ORGANIZATION BASED ON TEMPORAL DYNAMICS CHAPTER 6 PERCEPTUAL ORGANIZATION BASED ON TEMPORAL DYNAMICS This chapter presents a computational model for perceptual organization. A figure-ground segregation network is proposed based on a novel boundary

More information

Shape Modeling and Geometry Processing

Shape Modeling and Geometry Processing 252-0538-00L, Spring 2018 Shape Modeling and Geometry Processing Discrete Differential Geometry Differential Geometry Motivation Formalize geometric properties of shapes Roi Poranne # 2 Differential Geometry

More information

Geometric Mean Algorithms Based on Harmonic and Arithmetic Iterations

Geometric Mean Algorithms Based on Harmonic and Arithmetic Iterations Geometric Mean Algorithms Based on Harmonic and Arithmetic Iterations Ben Jeuris and Raf Vandebril KU Leuven, Dept. of Computer Science, 3001 Leuven(Heverlee), Belgium {ben.jeuris,raf.vandebril}@cs.kuleuven.be

More information

Denoising an Image by Denoising its Components in a Moving Frame

Denoising an Image by Denoising its Components in a Moving Frame Denoising an Image by Denoising its Components in a Moving Frame Gabriela Ghimpețeanu 1, Thomas Batard 1, Marcelo Bertalmío 1, and Stacey Levine 2 1 Universitat Pompeu Fabra, Spain 2 Duquesne University,

More information

Points Lines Connected points X-Y Scatter. X-Y Matrix Star Plot Histogram Box Plot. Bar Group Bar Stacked H-Bar Grouped H-Bar Stacked

Points Lines Connected points X-Y Scatter. X-Y Matrix Star Plot Histogram Box Plot. Bar Group Bar Stacked H-Bar Grouped H-Bar Stacked Plotting Menu: QCExpert Plotting Module graphs offers various tools for visualization of uni- and multivariate data. Settings and options in different types of graphs allow for modifications and customizations

More information

Geometric structures on manifolds

Geometric structures on manifolds CHAPTER 3 Geometric structures on manifolds In this chapter, we give our first examples of hyperbolic manifolds, combining ideas from the previous two chapters. 3.1. Geometric structures 3.1.1. Introductory

More information

Basic Algorithms for Digital Image Analysis: a course

Basic Algorithms for Digital Image Analysis: a course Institute of Informatics Eötvös Loránd University Budapest, Hungary Basic Algorithms for Digital Image Analysis: a course Dmitrij Csetverikov with help of Attila Lerch, Judit Verestóy, Zoltán Megyesi,

More information

274 Curves on Surfaces, Lecture 5

274 Curves on Surfaces, Lecture 5 274 Curves on Surfaces, Lecture 5 Dylan Thurston Notes by Qiaochu Yuan Fall 2012 5 Ideal polygons Previously we discussed three models of the hyperbolic plane: the Poincaré disk, the upper half-plane,

More information

3D Models and Matching

3D Models and Matching 3D Models and Matching representations for 3D object models particular matching techniques alignment-based systems appearance-based systems GC model of a screwdriver 1 3D Models Many different representations

More information

ECG782: Multidimensional Digital Signal Processing

ECG782: Multidimensional Digital Signal Processing Professor Brendan Morris, SEB 3216, brendan.morris@unlv.edu ECG782: Multidimensional Digital Signal Processing Spring 2014 TTh 14:30-15:45 CBC C313 Lecture 06 Image Structures 13/02/06 http://www.ee.unlv.edu/~b1morris/ecg782/

More information

Chapter 3. Bootstrap. 3.1 Introduction. 3.2 The general idea

Chapter 3. Bootstrap. 3.1 Introduction. 3.2 The general idea Chapter 3 Bootstrap 3.1 Introduction The estimation of parameters in probability distributions is a basic problem in statistics that one tends to encounter already during the very first course on the subject.

More information

Visual Recognition: Image Formation

Visual Recognition: Image Formation Visual Recognition: Image Formation Raquel Urtasun TTI Chicago Jan 5, 2012 Raquel Urtasun (TTI-C) Visual Recognition Jan 5, 2012 1 / 61 Today s lecture... Fundamentals of image formation You should know

More information

Processing of binary images

Processing of binary images Binary Image Processing Tuesday, 14/02/2017 ntonis rgyros e-mail: argyros@csd.uoc.gr 1 Today From gray level to binary images Processing of binary images Mathematical morphology 2 Computer Vision, Spring

More information

SUPPLEMENTARY FILE S1: 3D AIRWAY TUBE RECONSTRUCTION AND CELL-BASED MECHANICAL MODEL. RELATED TO FIGURE 1, FIGURE 7, AND STAR METHODS.

SUPPLEMENTARY FILE S1: 3D AIRWAY TUBE RECONSTRUCTION AND CELL-BASED MECHANICAL MODEL. RELATED TO FIGURE 1, FIGURE 7, AND STAR METHODS. SUPPLEMENTARY FILE S1: 3D AIRWAY TUBE RECONSTRUCTION AND CELL-BASED MECHANICAL MODEL. RELATED TO FIGURE 1, FIGURE 7, AND STAR METHODS. 1. 3D AIRWAY TUBE RECONSTRUCTION. RELATED TO FIGURE 1 AND STAR METHODS

More information

Image Coding with Active Appearance Models

Image Coding with Active Appearance Models Image Coding with Active Appearance Models Simon Baker, Iain Matthews, and Jeff Schneider CMU-RI-TR-03-13 The Robotics Institute Carnegie Mellon University Abstract Image coding is the task of representing

More information

CPSC 340: Machine Learning and Data Mining. Principal Component Analysis Fall 2016

CPSC 340: Machine Learning and Data Mining. Principal Component Analysis Fall 2016 CPSC 340: Machine Learning and Data Mining Principal Component Analysis Fall 2016 A2/Midterm: Admin Grades/solutions will be posted after class. Assignment 4: Posted, due November 14. Extra office hours:

More information

Particle Filtering. CS6240 Multimedia Analysis. Leow Wee Kheng. Department of Computer Science School of Computing National University of Singapore

Particle Filtering. CS6240 Multimedia Analysis. Leow Wee Kheng. Department of Computer Science School of Computing National University of Singapore Particle Filtering CS6240 Multimedia Analysis Leow Wee Kheng Department of Computer Science School of Computing National University of Singapore (CS6240) Particle Filtering 1 / 28 Introduction Introduction

More information

An optimization method for generating self-equilibrium shape of curved surface from developable surface

An optimization method for generating self-equilibrium shape of curved surface from developable surface 25-28th September, 2017, Hamburg, Germany Annette Bögle, Manfred Grohmann (eds.) An optimization method for generating self-equilibrium shape of curved surface from developable surface Jinglan CI *, Maoto

More information

Data Mining Chapter 3: Visualizing and Exploring Data Fall 2011 Ming Li Department of Computer Science and Technology Nanjing University

Data Mining Chapter 3: Visualizing and Exploring Data Fall 2011 Ming Li Department of Computer Science and Technology Nanjing University Data Mining Chapter 3: Visualizing and Exploring Data Fall 2011 Ming Li Department of Computer Science and Technology Nanjing University Exploratory data analysis tasks Examine the data, in search of structures

More information

CS 195-5: Machine Learning Problem Set 5

CS 195-5: Machine Learning Problem Set 5 CS 195-5: Machine Learning Problem Set 5 Douglas Lanman dlanman@brown.edu 26 November 26 1 Clustering and Vector Quantization Problem 1 Part 1: In this problem we will apply Vector Quantization (VQ) to

More information