Learning Two-Dimensional Shapes using Wavelet Local Extrema
|
|
- Theodore Maximillian Elliott
- 5 years ago
- Views:
Transcription
1 Learning Two-Dimensional Shapes using Wavelet Local Extrema tyuichi NAKAMURA t Institute of Information Sciences and Electronics University of Tsukuba Tsukuba City, 305, JAPAN STsu t omu Y 0 SHID A $ Communications Policy Bureau Ministry of Posts and Telecommunications Kasumigaseki, Chiyoda-ku, To kyo, , JAPAN Abstract The baszc zdea of the method for shape learnang as presented. Each shape as haerarchzcally characterazed wath sangularataes (local extrema) of wavelet transform of ats tangentaal oraentataon functaon. By takang the common structure of the sangulamtaes of two OT more shapes, the common characterzstzcs of the anput shapes are obtaaned. Through thas operatzon, not only the rough outlane of a shape but also several amportant characterastzcs of shapes are acquared. 1 Introduction Generally, a shape has a great amount of information. Often, we have to handle not only rough outline of a shape, but also detailed corners or 1D texture or irregular structure which have quite important information of a shape. Learning shapes, therefore, is a quite difficult problem. One of the difficulties arises from representation methods of a shape. Many of the early methods which analyze shapes axe based on approximation using polygons or other analytic curves. Because these methods destroy the detailed structures of an input shape, they can not be powerful tools for shape acquisition. On the contrary, there are methods based on multiscaling (especially, scale-space filtering method) in which behaviors of a signal over a wide range of spatial scale are analyzed. Most of all, scale-space filtering method has great advantages on the point that wide range of changes of curvature, that correspond to from rough outline of a shape to small projections, can be represented (e.g.[ab86, MM86, US90]). This method, however, has disadvantages on the completeness of the obtained data. It is proved that the fingerprint of zero-crossing points does not uniquely characterize the original function. In other words, while the convexity or concavity of frag- ments between inflection points is held, the quantity (or a detailed shape) of fragments is not preserved. Although some researches concentrated on the reconstruction problem [HM89, WenSO], enough results have not been achieved. For this purpose, we propose a new framework to deal with shape learning problem. Previously, we proposed the method for shape description by wavelet local extrema (that is local extrema extracted from wavelet transform of an input shape)[nsn93]. In this method, the detailed structure of a shape is fully characterized by wavelet local extrema. Several quite important characteristics of a shape can be quantized by the behavior of local extrema. This implies that these characteristics of shapes can be learned by using these extrema. Moreover, it is proved by Mallat [MZ92] that a close approximation of the original signal can be reconstructed from wavelet local extrema using alternate projections. This enables us to visualize (reconstruct) the acquired structures of shapes, which can be models or templates. In the followings, we will present the relationships between characteristics of shapes and wavelet local extrema, the basic ideaof this method, and the experiments of shape acquisition. 2 Wavelet Local Extrema and Characteristics of Shapes In our method, local extrema are extracted from wavelet transform. This is based on Mallat s signal characterization method, in which very close approximation of the original signal can be reconstructed from the local extrema. 2.1 Wavelet Transform and Wavelet Local Extrema Let baszc wavelet be $(t). Wavelet Function can be obtained by dilating the basic wavelet by a scaling factor S. The wavelet transform of f(t) is computed by convolving the signal with a dilated wavelet. Wdt) = f (t) * $s(t) (2) /94 $ IEEE 48
2 H H d f - WawI*1 Tmulorm... wsvdm Tnrulom...* Prolrtbn Op.mtor ~ 2 ) Figure 1: Wavelet Transform and Signal Reconstruction from Wavelet Local Extrema (a) Shape Contour - I I I I The wavelet transform that we use can be organized similar to quadratic mirror filter [EG77]. First, let us define a smoothing function 4(t) which satisfies the following formula. This formula means that each wavelet function is an independent component obtained by subtracting one smoothing function from the successive smoothing function. S2nf(~) is a smoothed signal by using &p(t)(= &(&)). It can be denoted as follows: %"f(w) = mi244 (4) We assume that the description at (j + 1)th level can be extracted from the description of jth level. For this purpose, we define low-pass filter fi(w). &+lf(w) = B(2jW)S2J(W) (5) Wavelet transform is considered as the mutually <independent component of the above description. To get wavelet transform we introduce high-pass filter G which satisfies = G(w)~(w). kv2r+lf(w) = G(2ju)S2J(w) (6) We can get wavelet transform by calculating W23f(t) changing j from 1 to n ( ma scale) according to the above formulae. While and G(w) are under constrained, the following set of filters satisfy the above conditions for any integer n (In the experiments, we set n = 0). W H(w) = ei"/2(cos( -))2n+l 2 G(w> = 4ieiw/2 sin(z) 2 (7) Fig. 1 shows a brief overview of our wavelet transform. The local extrema of wavelet transform can be extracted from each wavelet transform by simple operations (differentiation and extraction of zero-cross points). Mallat developed the method to reconstruct the close approximation of the original signal from these extrema. The method uses alternative convex projections between h(t) (a function which has the same local extremum values as those of original f(t)) and g(t) (a function ob- (b) 4(s) (c) Wavelet Local Extrema Figure 2: Example of Wavelet Local Extrema tained by inverse wavelet transform of a function which has finite energy). His experiments showed that iterations from 20 to 50 times are enough to get a very close approximation of the original signal. The details are skipped for lack of space (see [MZ92] for details). This method requires all wavelet local extrema and the smoothed function S2nf(~) at the largest scale. It can be shown, however, that if 2" is large enough, we can regard Spf(t) as a direct current component (that is, mean value of a signal). This implies that we only have to record one value or one extremum for Sznf(~). Thus we can reconstruct the original signal by using only wavelet local extrema on the condition that the number of levels is large enough. 2.2 Shape characteristics and Wavelet Local Extrema Generally, the contour of a two-dimensional shape can be expressed as one dimensional signal. For practical reasons, we expressed a shape by the tangential orientation 8(s) of each contour point, where s is the arc length from the starting point on the contour. This signal is transformed, then local extrema are extracted. These extrema hold complete information of the input shape. An example is shown in Fig. 2. The contour of China is shown in (a), and its tangential function (8(s)) in (b). The tangential function is wavelet transformed, then, local extrema are extracted. In Fig. 2(c), an example of a more simple shape is given, in which we can see that an extremum corresponds to a corner at each scale; positive extrema correspond to convex corners, and negative extrema to concave corners. The behaviors of extrema show the quite important characteristics of a curve. The detailed relationships between the characteristics of a shape and extrema are as follows: Extremum Position: Position of a corner at each scale. In Fig. 2(c), corners (a--f) generate extrema 49
3 CI position angle sharpness compound structure Figure 3: The basic idea of corner interpolation in the graph below. Extremum Value: Changing angle of a corner '. In Fig. 2(c), a corner with an acute angle (b) makes large extremum value, while a corner with obtuse angle (a) makes small extremum value. Value Changes across scales: Shape of a corner. If the value increases as the scale decreases, the discontinuity of a corner is large. In Fig. 2(c), a sharp corner (f) makes a series of extrema that the value increases as the scale decreases, while a dull corner (d) makes those that value decreases. This phenomenon can be explained by the relationship between Lipschitz exponents and the decay of wavelet transform [MH92]. Position Shift across scales: A combination of corners. If the position shifts as scale changes, this corner should be considered as a portion of a compound corner. Combination of adjacent extrema: A combination of extrema expresses the compoud structure of a corner. 2.3 Learning Shape Characteristics Using these relationships, we can consider several pseudo axes of corner characteristics which are position, angle, and sharpness (singularity or discontinuity). Although it is still difficult to directly compose a corner which has desired characteristics, interpolation and extrapolation of their characteristics are usually easy to acquire. The basic idea is shown in Fig. 3. For example, when the positions of the extrema are interpolated, the positions of the corners are interpolated. Although there is no proof that a real shape exits which have extrema with these positions and values, a similar shape (which has extrema with close positions and close values) is almost always obtained. The reason is the stability of Mallat's reconstruction method which uses convex projection. The flow of shape learning process in our method is as follows: 'This is only a rough relationship. Strictly speaking, the value of envelope should be considered at the same time. Wavelet transform and extrema extraction Finding correspondences among extrema (a) hierarchical organization within a shape (b) hierarchical matching between shapes Generation of new extrema and reconstruction to a shape. In the following sections, 2 and 3 are described. 3 Finding Correspondences by Shape Matching To find correspondences of corners across shapes, we need shape matching. First, relationships among extrema within a shape are searched. The operation is quite similar to the process of getting fingerprints for the scale-space expression. Then, matching between shapes is performed. Since we have already shown the method for this operation in [NSN93], a brief overview is given in the followings. Note that we do not use positive minima and negative maxima for simplicity because they are not necessary for the shape reconstruction (However, they are important information as connecting points between convex corners or between concave corners). 3.1 Getting Hierarchy Dynamic Programming is used to get the correspondences between extrema across scales. Although the basic idea is similar to [US90], the details are different because the nodes used in this research are wavelet local extrema, First, let us denote two sets of nodes (hereafter we call data structure for an extremum node) at two consecutive scales as follows: (9) where pi" is a i-th node (extrema) at k-th level, and Pk is a set of extrema at k-th level. Each of them includes all nodes sorted according to their position at their scale. This matching can be considered as an operation to find mapping function J which minimizes the following formula: where, s(i, J(i)) is a score of matching between two nodes. For this matching, we placed a few constraints on the correspondences across scales. The penalty values (the costs of paths) of DP matching is determined according to the following conditions: 1. pf" and p;(i) have the same sign. 50
4 (a) Chinal (c) China2 (b) Hierarchy within Chinal (d) Hierarchy within China2 Nodes express the Wavelet Extrema for each shape. The extrema at the same scale are horizontally aligned, while those at different scales are vertically aligned. An arc expresses the relationship between a parent and child. Figure 4: Example of two shapes 2. The positions of two nodes are close enough to each other. 3. The extremum value and the of two nodes are close to each other. In this step, matching is recursively performed from the largest scale (P" and P-') to the smallest scale (P2 and P'). An example is shown in Fig. 4, in which extrema from two shapes are hierarchically organized. 3.2 Shape Matching by DP method We also applied DP matching to this problem. The obtained hierarchy of wavelet local extrema within a shape is used for this purpose. The type, value, and position of extrema can be considered as conditions for matching. As mentioned in the previous section, this process is also performed from coarse to fine resolution. Let us denote the two sets of nodes at k-th scale as follows: P1h ={pl;li=o,l,...,n'} (11) Pa'" = (p2;jj = 0,1,..., N ~ ) (12) The objective of shape matching is to find J which minimizes the following formula: N, -1 where s(i,j) is the penalty value of the correspondence between pl: and p2;. The penalty values of this matching are determined according to the following conditions: scale 27 scale 25 scale 23 scale 2' Figure 5: Matching Result 1. Maxima must match maxima, and minima must match minima. 2. The extremum values should be close to each other. 3. The distances from the previous extremum (or previously matched extremum) should be close to each other. 4. The match of parents nodes at larger scale are used as constraints. Assume that plt have a hierarchical correspondence to plr+', and p2fc have a hierarchical correspondence to p2&+'. If p2&+' does not match plq'l, plt has less possibility of correspondence to p2;. The result of matching at each scale is shown in Fig. 5. Extrema are placed on the contour, and corresponding extrema are related by arcs (Each shape is reconstructed by using wavelet transform at its scale). 4 Learning Shape 4.1 Learning Corner Characteristics If the correspondences of corners are correctly determined, we can get interpolation of a corner by simply interpolating extrema values along scales. The operation can be simply denoted as follows: U, = k.?i1 + (1- k) * U2 (14) where?i is either the value or the position of a extremum. For example, zli means the extremum value of shapel. In this formula, 0 L. k L. 1 means interpolation, otherwise extrapolation. A simple example is shown in Fig. 6. In this example, each of two original shapes (originall and original2) has a concave corner with different characteristics near the center of each shape. The shape in (d) is obtained when k = 0.5. By averaging extrema which have correspondences, we can get quite good approximation for both shapes. For (c) and (e), we set k = 1.5 and k = -0.5 respectively. The obtained shapes show that their characteristics are well interpolated or extrapolated. In our experiments, interporation works fairly well, while extrapolation sometimes gives quite distorted results2. 'There is no guarantee that a shape exists which has such characteristics as the extrapolations of those of input shapes! 51
5 (a) Acquired shape (b) Extrapolation (c) Simple average Figure 7: Acquired Shapes (c) extrapolation (d) interpolation (e) extrapolation (over 1) (over 2) Figure 6: Interpolation of Corners If we consider three or more shapes as input, the notions of interpolation and extrapolation get inappropriate. In this case, we can think of averaging and emphasizing differences from the average. ij=- cy=^ vi n 2rne, = k4+(1 -k).va (15) where V means the average of extrema taken from input shapes. 4.2 Learning Whole Shape Given a shape matching result, we have the extrema correspondences for an entire shape. We can consider these corresponding extrema as the common structure of two shapes. Learning the whole shape structure, therefore, is achieved by applying the method in the previous section to all the pair of corresponding extrema. In other words, the interpolation of whole shape can be obtained by interpolating all pairs of corresponding extrema. An example is shown in Fig. 7. The shape reconstructed from all the interpolated extrema is shown. It shows that the rough outline, which is common in two shapes, is correctly obtained. In addition to that, the characteristics of the corners in the original shapes are preserved well in the acquired shape. (However, the upper part of the reconstructed shape is slightly distorted because the shape matching is not completely succeeded for that part). For comparison, the obtained shape by taking simple average of xy-coordinates of two shapes is shown as (c). Although the rough outline is also preserved in this shape, we can see a lot of small projections which belongs to only one of the shapes, and upper part of the shape is distorted. An example of extrapolation is also shown in Fig. 7. We can see the differences between two original shapes are emphasized by the ext,rapolation. An example for three or more shapes is shown in Fig. 8. Using the shape matching results between Fishl and each of other shapes of fish (Fishl vs. Fish2, Fishl vs. Fish3,...), we obtained the average by the formula Fishl Fish2 Fish3 Fish4 Fish5 Acquired Shape Figure 8: Example of Shape Acquisition from five shapes (15). In this example, the characteristics of dorsal fins and pectral fins are acquired quite well, since all of the shapes of fish have simlar structures around them. Ventral or anal fin is distorted because input shapes do not have common structure around it. 5 Summary This paper proposed a new framework for shape learning using wavelet local extrema. In this method, not only the rough outline of a shape but also several characteristics of a shape are preserved. The basic idea of this method is simple enough to be applied to many kinds of shape acquisition. Therefore, it can be a new basis for bidirectional process of shapes; analysis, learning, and generation. References [AB861 [EG77] [H h4 891 [MH92] [MM861 [MZ92] [NSN93] [US90] [Wen 9 01 H. Asada and M. Brady. Curvature Primal Sketch. IEEE PAMI, Vol. 8, No. 1, pp. 2-14, Jan D. Esteban and C. Galand. Application of quadrature mirror filters to split band voice coding schemes. Proc. Int. Conj. ASSP, pp , R. Hummel and R. Moniot. Reconstructions from Zero Crossings in Scale Space. IEEE ASSP, Vol. 37, No. 12, pp , dec S. Mallat and W. Hwang. Singularity Detection and Processing with Wavelets. IEEE Inform. Theory, Vol. 38, No. 2, pp , March F. Mokhtarian and A. Mackworth. Scale-Based Description and Recognition of Planar Curves and Two-Dimensional Shapes. IEEE PAMI, Vol. 8, No. 1, pp , Jan S. Mallat and S. Zhong. Characterization of Signals from Multiscale Edges. IEEE PAMI, Vol. 14, No. 7, pp , jul Y. Nakamura. K. Satoda, and M. Nagao. Shape description and matching using wavelet extrema. Proc. Asian Conference on Computer Vasion, pp , N. Ueda and S. Suzuki. Automatic shape model acquisition using multiscale segment matching. Proc. foth ICPR, Vol. 1, pp , J. Weng. A Theory of Image Matching. Proc. Yrd ICCV, pp ,
A Modified Spline Interpolation Method for Function Reconstruction from Its Zero-Crossings
Scientific Papers, University of Latvia, 2010. Vol. 756 Computer Science and Information Technologies 207 220 P. A Modified Spline Interpolation Method for Function Reconstruction from Its Zero-Crossings
More informationRegularity Analysis of Non Uniform Data
Regularity Analysis of Non Uniform Data Christine Potier and Christine Vercken Abstract. A particular class of wavelet, derivatives of B-splines, leads to fast and ecient algorithms for contours detection
More informationA New Approach to Computation of Curvature Scale Space Image for Shape Similarity Retrieval
A New Approach to Computation of Curvature Scale Space Image for Shape Similarity Retrieval Farzin Mokhtarian, Sadegh Abbasi and Josef Kittler Centre for Vision Speech and Signal Processing Department
More informationA New Technique of Extraction of Edge Detection Using Digital Image Processing
International OPEN ACCESS Journal Of Modern Engineering Research (IJMER) A New Technique of Extraction of Edge Detection Using Digital Image Processing Balaji S.C.K 1 1, Asst Professor S.V.I.T Abstract:
More informationBoundary descriptors. Representation REPRESENTATION & DESCRIPTION. Descriptors. Moore boundary tracking
Representation REPRESENTATION & DESCRIPTION After image segmentation the resulting collection of regions is usually represented and described in a form suitable for higher level processing. Most important
More informationDirect Rendering. Direct Rendering Goals
May 2, 2005 Goals General Goals Small memory footprint Fast rendering High-quality results identical to those of Saffron V1 using distance-based anti-aliasing and alignment zones Goals Specific Goals Avoid
More informationAffine-invariant shape matching and recognition under partial occlusion
Title Affine-invariant shape matching and recognition under partial occlusion Author(s) Mai, F; Chang, CQ; Hung, YS Citation The 17th IEEE International Conference on Image Processing (ICIP 2010), Hong
More informationAn Image Curvature Microscope
An Jean-Michel MOREL Joint work with Adina CIOMAGA and Pascal MONASSE Centre de Mathématiques et de Leurs Applications, Ecole Normale Supérieure de Cachan Séminaire Jean Serra - 70 ans April 2, 2010 Jean-Michel
More informationLet s start with occluding contours (or interior and exterior silhouettes), and look at image-space algorithms. A very simple technique is to render
1 There are two major classes of algorithms for extracting most kinds of lines from 3D meshes. First, there are image-space algorithms that render something (such as a depth map or cosine-shaded model),
More informationVoronoi Diagram. Xiao-Ming Fu
Voronoi Diagram Xiao-Ming Fu Outlines Introduction Post Office Problem Voronoi Diagram Duality: Delaunay triangulation Centroidal Voronoi tessellations (CVT) Definition Applications Algorithms Outlines
More informationa) y = x 3 + 3x 2 2 b) = UNIT 4 CURVE SKETCHING 4.1 INCREASING AND DECREASING FUNCTIONS
UNIT 4 CURVE SKETCHING 4.1 INCREASING AND DECREASING FUNCTIONS We read graphs as we read sentences: left to right. Plainly speaking, as we scan the function from left to right, the function is said to
More informationLECTURE 18 - OPTIMIZATION
LECTURE 18 - OPTIMIZATION CHRIS JOHNSON Abstract. In this lecture we ll describe extend the optimization techniques you learned in your first semester calculus class to optimize functions of multiple variables.
More informationWATERSHED, HIERARCHICAL SEGMENTATION AND WATERFALL ALGORITHM
WATERSHED, HIERARCHICAL SEGMENTATION AND WATERFALL ALGORITHM Serge Beucher Ecole des Mines de Paris, Centre de Morphologie Math«ematique, 35, rue Saint-Honor«e, F 77305 Fontainebleau Cedex, France Abstract
More informationDOWNLOAD PDF BIG IDEAS MATH VERTICAL SHRINK OF A PARABOLA
Chapter 1 : BioMath: Transformation of Graphs Use the results in part (a) to identify the vertex of the parabola. c. Find a vertical line on your graph paper so that when you fold the paper, the left portion
More informationLevel lines based disocclusion
Level lines based disocclusion Simon Masnou Jean-Michel Morel CEREMADE CMLA Université Paris-IX Dauphine Ecole Normale Supérieure de Cachan 75775 Paris Cedex 16, France 94235 Cachan Cedex, France Abstract
More informationEnhanced Active Shape Models with Global Texture Constraints for Image Analysis
Enhanced Active Shape Models with Global Texture Constraints for Image Analysis Shiguang Shan, Wen Gao, Wei Wang, Debin Zhao, Baocai Yin Institute of Computing Technology, Chinese Academy of Sciences,
More informationMath 1020 Objectives & Exercises Calculus Concepts Spring 2019
Section of Textbook 1.1 AND Learning Objectives/Testable Skills Identify four representations of a function. Specify input and output variables, input and output descriptions, and input and output units.
More informationcoding of various parts showing different features, the possibility of rotation or of hiding covering parts of the object's surface to gain an insight
Three-Dimensional Object Reconstruction from Layered Spatial Data Michael Dangl and Robert Sablatnig Vienna University of Technology, Institute of Computer Aided Automation, Pattern Recognition and Image
More informationRipplet: a New Transform for Feature Extraction and Image Representation
Ripplet: a New Transform for Feature Extraction and Image Representation Dr. Dapeng Oliver Wu Joint work with Jun Xu Department of Electrical and Computer Engineering University of Florida Outline Motivation
More informationEdge and local feature detection - 2. Importance of edge detection in computer vision
Edge and local feature detection Gradient based edge detection Edge detection by function fitting Second derivative edge detectors Edge linking and the construction of the chain graph Edge and local feature
More informationChapter 3 Image Registration. Chapter 3 Image Registration
Chapter 3 Image Registration Distributed Algorithms for Introduction (1) Definition: Image Registration Input: 2 images of the same scene but taken from different perspectives Goal: Identify transformation
More informationAdaptive-Mesh-Refinement Pattern
Adaptive-Mesh-Refinement Pattern I. Problem Data-parallelism is exposed on a geometric mesh structure (either irregular or regular), where each point iteratively communicates with nearby neighboring points
More informationDETECTION OF CHANGES IN SURVEILLANCE VIDEOS. Longin Jan Latecki, Xiangdong Wen, and Nilesh Ghubade
DETECTION OF CHANGES IN SURVEILLANCE VIDEOS Longin Jan Latecki, Xiangdong Wen, and Nilesh Ghubade CIS Dept. Dept. of Mathematics CIS Dept. Temple University Temple University Temple University Philadelphia,
More informationFACET SHIFT ALGORITHM BASED ON MINIMAL DISTANCE IN SIMPLIFICATION OF BUILDINGS WITH PARALLEL STRUCTURE
FACET SHIFT ALGORITHM BASED ON MINIMAL DISTANCE IN SIMPLIFICATION OF BUILDINGS WITH PARALLEL STRUCTURE GE Lei, WU Fang, QIAN Haizhong, ZHAI Renjian Institute of Surveying and Mapping Information Engineering
More informationLearning and Inferring Depth from Monocular Images. Jiyan Pan April 1, 2009
Learning and Inferring Depth from Monocular Images Jiyan Pan April 1, 2009 Traditional ways of inferring depth Binocular disparity Structure from motion Defocus Given a single monocular image, how to infer
More informationTexture Boundary Detection - A Structural Approach
Texture Boundary Detection - A Structural Approach Wen Wen Richard J. Fryer Machine Perception Research Group, Department of Computer Science University of Strathclyde, Glasgow Gl 1XH, United Kingdom Abstract
More information2D Spline Curves. CS 4620 Lecture 18
2D Spline Curves CS 4620 Lecture 18 2014 Steve Marschner 1 Motivation: smoothness In many applications we need smooth shapes that is, without discontinuities So far we can make things with corners (lines,
More informationMTAEA Convexity and Quasiconvexity
School of Economics, Australian National University February 19, 2010 Convex Combinations and Convex Sets. Definition. Given any finite collection of points x 1,..., x m R n, a point z R n is said to be
More informationTopic 6 Representation and Description
Topic 6 Representation and Description Background Segmentation divides the image into regions Each region should be represented and described in a form suitable for further processing/decision-making Representation
More informationCOMPUTER AND ROBOT VISION
VOLUME COMPUTER AND ROBOT VISION Robert M. Haralick University of Washington Linda G. Shapiro University of Washington A^ ADDISON-WESLEY PUBLISHING COMPANY Reading, Massachusetts Menlo Park, California
More informationCS 534: Computer Vision Texture
CS 534: Computer Vision Texture Ahmed Elgammal Dept of Computer Science CS 534 Texture - 1 Outlines Finding templates by convolution What is Texture Co-occurrence matrices for texture Spatial Filtering
More informationCompression of 3D Objects with Multistage Color-Depth Panoramic Maps
Compression of 3D Objects with Multistage Color-Depth Panoramic Maps Chang-Ming Tsai 1,2, Wen-Yan Chang 1, Chu-Song Chen 1, Gregory Y. Tang 2 1 Institute of Information Science, Academia Sinica, Nankang
More informationCS 664 Segmentation. Daniel Huttenlocher
CS 664 Segmentation Daniel Huttenlocher Grouping Perceptual Organization Structural relationships between tokens Parallelism, symmetry, alignment Similarity of token properties Often strong psychophysical
More informationA decision support system for ship identification based on the curvature scale space representation
A decision support system for ship identification based on the curvature scale space representation Álvaro Enríquez de Luna 1, Carlos Miravet 2, Deitze Otaduy 1, Carlos Dorronsoro 1 1 Centro de Investigación
More informationRobust Video Super-Resolution with Registration Efficiency Adaptation
Robust Video Super-Resolution with Registration Efficiency Adaptation Xinfeng Zhang a, Ruiqin Xiong b, Siwei Ma b, Li Zhang b, Wen Gao b a Institute of Computing Technology, Chinese Academy of Sciences,
More informationFingerprint Image Compression
Fingerprint Image Compression Ms.Mansi Kambli 1*,Ms.Shalini Bhatia 2 * Student 1*, Professor 2 * Thadomal Shahani Engineering College * 1,2 Abstract Modified Set Partitioning in Hierarchical Tree with
More informationLecture 10: Image Descriptors and Representation
I2200: Digital Image processing Lecture 10: Image Descriptors and Representation Prof. YingLi Tian Nov. 15, 2017 Department of Electrical Engineering The City College of New York The City University of
More informationHOUGH TRANSFORM CS 6350 C V
HOUGH TRANSFORM CS 6350 C V HOUGH TRANSFORM The problem: Given a set of points in 2-D, find if a sub-set of these points, fall on a LINE. Hough Transform One powerful global method for detecting edges
More informationParameterization. Michael S. Floater. November 10, 2011
Parameterization Michael S. Floater November 10, 2011 Triangular meshes are often used to represent surfaces, at least initially, one reason being that meshes are relatively easy to generate from point
More informationData Partitioning. Figure 1-31: Communication Topologies. Regular Partitions
Data In single-program multiple-data (SPMD) parallel programs, global data is partitioned, with a portion of the data assigned to each processing node. Issues relevant to choosing a partitioning strategy
More informationA SYNTAX FOR IMAGE UNDERSTANDING
A SYNTAX FOR IMAGE UNDERSTANDING Narendra Ahuja University of Illinois at Urbana-Champaign May 21, 2009 Work Done with. Sinisa Todorovic, Mark Tabb, Himanshu Arora, Varsha. Hedau, Bernard Ghanem, Tim Cheng.
More informationAutoregressive and Random Field Texture Models
1 Autoregressive and Random Field Texture Models Wei-Ta Chu 2008/11/6 Random Field 2 Think of a textured image as a 2D array of random numbers. The pixel intensity at each location is a random variable.
More informationEE795: Computer Vision and Intelligent Systems
EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 WRI C225 Lecture 04 130131 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Review Histogram Equalization Image Filtering Linear
More informationHP ICVGIP competition
HP ICVGIP competition Indian Institute of Technology KHARAGPUR Project title Automated Mosaicing of Torn Paper Documents By Amiya Patanaik Bibek Behera Sukadeb Acharya Project guide Dr. Abhir Bhalerao
More informationLocal Image Registration: An Adaptive Filtering Framework
Local Image Registration: An Adaptive Filtering Framework Gulcin Caner a,a.murattekalp a,b, Gaurav Sharma a and Wendi Heinzelman a a Electrical and Computer Engineering Dept.,University of Rochester, Rochester,
More informationUNIVERSITY OF OSLO. Faculty of Mathematics and Natural Sciences
UNIVERSITY OF OSLO Faculty of Mathematics and Natural Sciences Exam: INF 4300 / INF 9305 Digital image analysis Date: Thursday December 21, 2017 Exam hours: 09.00-13.00 (4 hours) Number of pages: 8 pages
More informationCHAPTER 6 DETECTION OF MASS USING NOVEL SEGMENTATION, GLCM AND NEURAL NETWORKS
130 CHAPTER 6 DETECTION OF MASS USING NOVEL SEGMENTATION, GLCM AND NEURAL NETWORKS A mass is defined as a space-occupying lesion seen in more than one projection and it is described by its shapes and margin
More informationFree-Hand Stroke Approximation for Intelligent Sketching Systems
MIT Student Oxygen Workshop, 2001. Free-Hand Stroke Approximation for Intelligent Sketching Systems Tevfik Metin Sezgin MIT Artificial Intelligence Laboratory mtsezgin@ai.mit.edu Abstract. Free-hand sketching
More informationIN5520 Digital Image Analysis. Two old exams. Practical information for any written exam Exam 4300/9305, Fritz Albregtsen
IN5520 Digital Image Analysis Two old exams Practical information for any written exam Exam 4300/9305, 2016 Exam 4300/9305, 2017 Fritz Albregtsen 27.11.2018 F13 27.11.18 IN 5520 1 Practical information
More informationOhio Tutorials are designed specifically for the Ohio Learning Standards to prepare students for the Ohio State Tests and end-ofcourse
Tutorial Outline Ohio Tutorials are designed specifically for the Ohio Learning Standards to prepare students for the Ohio State Tests and end-ofcourse exams. Math Tutorials offer targeted instruction,
More informationPetrel TIPS&TRICKS from SCM
Petrel TIPS&TRICKS from SCM Knowledge Worth Sharing Merging Overlapping Files into One 2D Grid Often several files (grids or data) covering adjacent and overlapping areas must be combined into one 2D Grid.
More informationCSE 417 Network Flows (pt 4) Min Cost Flows
CSE 417 Network Flows (pt 4) Min Cost Flows Reminders > HW6 is due Monday Review of last three lectures > Defined the maximum flow problem find the feasible flow of maximum value flow is feasible if it
More informationBLOCK-BASED discrete cosine transform (BDCT) [1]
450 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 14, NO. 4, APRIL 2004 Blocking Artifacts Suppression in Block-Coded Images Using Overcomplete Wavelet Representation Alan W.-C.
More informationShape Modeling. Differential Geometry Primer Smooth Definitions Discrete Theory in a Nutshell. CS 523: Computer Graphics, Spring 2011
CS 523: Computer Graphics, Spring 2011 Shape Modeling Differential Geometry Primer Smooth Definitions Discrete Theory in a Nutshell 2/15/2011 1 Motivation Geometry processing: understand geometric characteristics,
More informationADVANCED IMAGE PROCESSING METHODS FOR ULTRASONIC NDE RESEARCH C. H. Chen, University of Massachusetts Dartmouth, N.
ADVANCED IMAGE PROCESSING METHODS FOR ULTRASONIC NDE RESEARCH C. H. Chen, University of Massachusetts Dartmouth, N. Dartmouth, MA USA Abstract: The significant progress in ultrasonic NDE systems has now
More informationDEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING DS7201 ADVANCED DIGITAL IMAGE PROCESSING II M.E (C.S) QUESTION BANK UNIT I 1. Write the differences between photopic and scotopic vision? 2. What
More informationThe Farthest Point Delaunay Triangulation Minimizes Angles
The Farthest Point Delaunay Triangulation Minimizes Angles David Eppstein Department of Information and Computer Science UC Irvine, CA 92717 November 20, 1990 Abstract We show that the planar dual to the
More informationA The left scanline The right scanline
Dense Disparity Estimation via Global and Local Matching Chun-Jen Tsai and Aggelos K. Katsaggelos Electrical and Computer Engineering Northwestern University Evanston, IL 60208-3118, USA E-mail: tsai@ece.nwu.edu,
More information1.1 Pearson Modeling and Equation Solving
Date:. Pearson Modeling and Equation Solving Syllabus Objective:. The student will solve problems using the algebra of functions. Modeling a Function: Numerical (data table) Algebraic (equation) Graphical
More informationComputer Vision I. Announcements. Fourier Tansform. Efficient Implementation. Edge and Corner Detection. CSE252A Lecture 13.
Announcements Edge and Corner Detection HW3 assigned CSE252A Lecture 13 Efficient Implementation Both, the Box filter and the Gaussian filter are separable: First convolve each row of input image I with
More informationKey words: B- Spline filters, filter banks, sub band coding, Pre processing, Image Averaging IJSER
International Journal of Scientific & Engineering Research, Volume 7, Issue 9, September-2016 470 Analyzing Low Bit Rate Image Compression Using Filters and Pre Filtering PNV ABHISHEK 1, U VINOD KUMAR
More informationAn Automatic Hole Filling Method of Point Cloud for 3D Scanning
An Automatic Hole Filling Method of Point Cloud for 3D Scanning Yuta MURAKI Osaka Institute of Technology Osaka, Japan yuta.muraki@oit.ac.jp Koji NISHIO Osaka Institute of Technology Osaka, Japan koji.a.nishio@oit.ac.jp
More informationBirkdale High School - Higher Scheme of Work
Birkdale High School - Higher Scheme of Work Module 1 - Integers and Decimals Understand and order integers (assumed) Use brackets and hierarchy of operations (BODMAS) Add, subtract, multiply and divide
More informationA MORPHOLOGY-BASED FILTER STRUCTURE FOR EDGE-ENHANCING SMOOTHING
Proceedings of the 1994 IEEE International Conference on Image Processing (ICIP-94), pp. 530-534. (Austin, Texas, 13-16 November 1994.) A MORPHOLOGY-BASED FILTER STRUCTURE FOR EDGE-ENHANCING SMOOTHING
More informationA Singular Example for the Averaged Mean Curvature Flow
To appear in Experimental Mathematics Preprint Vol. No. () pp. 3 7 February 9, A Singular Example for the Averaged Mean Curvature Flow Uwe F. Mayer Abstract An embedded curve is presented which under numerical
More informationImage 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 informationScaled representations
Scaled representations Big bars (resp. spots, hands, etc.) and little bars are both interesting Stripes and hairs, say Inefficient to detect big bars with big filters And there is superfluous detail in
More information2D Spline Curves. CS 4620 Lecture 13
2D Spline Curves CS 4620 Lecture 13 2008 Steve Marschner 1 Motivation: smoothness In many applications we need smooth shapes [Boeing] that is, without discontinuities So far we can make things with corners
More informationPrentice Hall Mathematics: Geometry 2007 Correlated to: Arizona Academic Standards for Mathematics (Grades 9-12)
Strand 1: Number Sense and Operations Every student should understand and use all concepts and skills from the previous grade levels. The standards are designed so that new learning builds on preceding
More informationScientific Visualization Example exam questions with commented answers
Scientific Visualization Example exam questions with commented answers The theoretical part of this course is evaluated by means of a multiple- choice exam. The questions cover the material mentioned during
More informationComputer Vision I. Announcement. Corners. Edges. Numerical Derivatives f(x) Edge and Corner Detection. CSE252A Lecture 11
Announcement Edge and Corner Detection Slides are posted HW due Friday CSE5A Lecture 11 Edges Corners Edge is Where Change Occurs: 1-D Change is measured by derivative in 1D Numerical Derivatives f(x)
More informationComparative Analysis of Image Compression Using Wavelet and Ridgelet Transform
Comparative Analysis of Image Compression Using Wavelet and Ridgelet Transform Thaarini.P 1, Thiyagarajan.J 2 PG Student, Department of EEE, K.S.R College of Engineering, Thiruchengode, Tamil Nadu, India
More informationGeneralized Fuzzy Clustering Model with Fuzzy C-Means
Generalized Fuzzy Clustering Model with Fuzzy C-Means Hong Jiang 1 1 Computer Science and Engineering, University of South Carolina, Columbia, SC 29208, US jiangh@cse.sc.edu http://www.cse.sc.edu/~jiangh/
More informationSUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS
SUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS Cognitive Robotics Original: David G. Lowe, 004 Summary: Coen van Leeuwen, s1460919 Abstract: This article presents a method to extract
More informationEdge detection. Convert a 2D image into a set of curves. Extracts salient features of the scene More compact than pixels
Edge Detection Edge detection Convert a 2D image into a set of curves Extracts salient features of the scene More compact than pixels Origin of Edges surface normal discontinuity depth discontinuity surface
More informationReflector profile optimisation using Radiance
Reflector profile optimisation using Radiance 1,4 1,2 1, 8 6 4 2 3. 2.5 2. 1.5 1..5 I csf(1) csf(2). 1 2 3 4 5 6 Giulio ANTONUTTO Krzysztof WANDACHOWICZ page 1 The idea Krzysztof WANDACHOWICZ Giulio ANTONUTTO
More informationSeparation of Surface Roughness Profile from Raw Contour based on Empirical Mode Decomposition Shoubin LIU 1, a*, Hui ZHANG 2, b
International Conference on Advances in Mechanical Engineering and Industrial Informatics (AMEII 2015) Separation of Surface Roughness Profile from Raw Contour based on Empirical Mode Decomposition Shoubin
More informationTHE GEOMETRIC HEAT EQUATION AND SURFACE FAIRING
THE GEOMETRIC HEAT EQUATION AN SURFACE FAIRING ANREW WILLIS BROWN UNIVERSITY, IVISION OF ENGINEERING, PROVIENCE, RI 02912, USA 1. INTROUCTION This paper concentrates on analysis and discussion of the heat
More informationA biometric iris recognition system based on principal components analysis, genetic algorithms and cosine-distance
Safety and Security Engineering VI 203 A biometric iris recognition system based on principal components analysis, genetic algorithms and cosine-distance V. Nosso 1, F. Garzia 1,2 & R. Cusani 1 1 Department
More informationDerivatives and Graphs of Functions
Derivatives and Graphs of Functions September 8, 2014 2.2 Second Derivatives, Concavity, and Graphs In the previous section, we discussed how our derivatives can be used to obtain useful information about
More informationGraphs of Exponential
Graphs of Exponential Functions By: OpenStaxCollege As we discussed in the previous section, exponential functions are used for many realworld applications such as finance, forensics, computer science,
More information3D Shape Registration using Regularized Medial Scaffolds
3D Shape Registration using Regularized Medial Scaffolds 3DPVT 2004 Thessaloniki, Greece Sep. 6-9, 2004 Ming-Ching Chang Frederic F. Leymarie Benjamin B. Kimia LEMS, Division of Engineering, Brown University
More informationGeostatistics 2D GMS 7.0 TUTORIALS. 1 Introduction. 1.1 Contents
GMS 7.0 TUTORIALS 1 Introduction Two-dimensional geostatistics (interpolation) can be performed in GMS using the 2D Scatter Point module. The module is used to interpolate from sets of 2D scatter points
More informationMulti-Scale Free-Form Surface Description
Multi-Scale Free-Form Surface Description Farzin Mokhtarian, Nasser Khalili and Peter Yuen Centre for Vision Speech and Signal Processing Dept. of Electronic and Electrical Engineering University of Surrey,
More informationNonrigid Surface Modelling. and Fast Recovery. Department of Computer Science and Engineering. Committee: Prof. Leo J. Jia and Prof. K. H.
Nonrigid Surface Modelling and Fast Recovery Zhu Jianke Supervisor: Prof. Michael R. Lyu Committee: Prof. Leo J. Jia and Prof. K. H. Wong Department of Computer Science and Engineering May 11, 2007 1 2
More informationMath Exam 2a. 1) Take the derivatives of the following. DO NOT SIMPLIFY! 2 c) y = tan(sec2 x) ) b) y= , for x 2.
Math 111 - Exam 2a 1) Take the derivatives of the following. DO NOT SIMPLIFY! a) y = ( + 1 2 x ) (sin(2x) - x- x 1 ) b) y= 2 x + 1 c) y = tan(sec2 x) 2) Find the following derivatives a) Find dy given
More informationMöbius Transformations in Scientific Computing. David Eppstein
Möbius Transformations in Scientific Computing David Eppstein Univ. of California, Irvine School of Information and Computer Science (including joint work with Marshall Bern from WADS 01 and SODA 03) Outline
More informationGRADE 6 PAT REVIEW. Math Vocabulary NAME:
GRADE 6 PAT REVIEW Math Vocabulary NAME: Estimate Round Number Concepts An approximate or rough calculation, often based on rounding. Change a number to a more convenient value. (0 4: place value stays
More informationAn Image Curvature Microscope
An Jean-Michel MOREL Joint work with Adina CIOMAGA and Pascal MONASSE Centre de Mathématiques et de Leurs Applications, Ecole Normale Supérieure de Cachan Séminaire Jean Serra - 70 ans April 2, 2010 Jean-Michel
More informationXiaolin Li and Dit-Yan Yeung. Hong Kong University of Science and Technology. Clear Water Bay, Kowloon, HONG KONG
On-line Handwritten Alphanumeric Character Recognition Using Feature Sequences Xiaolin Li and Dit-Yan Yeung Department of Computer Science Hong Kong University of Science and Technology Clear Water Bay,
More informationCS334: Digital Imaging and Multimedia Edges and Contours. Ahmed Elgammal Dept. of Computer Science Rutgers University
CS334: Digital Imaging and Multimedia Edges and Contours Ahmed Elgammal Dept. of Computer Science Rutgers University Outlines What makes an edge? Gradient-based edge detection Edge Operators From Edges
More informationPyramid Coding and Subband Coding
Pyramid Coding and Subband Coding Predictive pyramids Transform pyramids Subband coding Perfect reconstruction filter banks Quadrature mirror filter banks Octave band splitting Transform coding as a special
More informationReducing Points In a Handwritten Curve (Improvement in a Note-taking Tool)
Reducing Points In a Handwritten Curve (Improvement in a Note-taking Tool) Kaoru Oka oka@oz.ces.kyutech.ac.jp Faculty of Computer Science and Systems Engineering Kyushu Institute of Technology Japan Ryoji
More informationVisibility: Finding the Staircase Kernel in Orthogonal Polygons
Visibility: Finding the Staircase Kernel in Orthogonal Polygons 8 Visibility: Finding the Staircase Kernel in Orthogonal Polygons Tzvetalin S. Vassilev, Nipissing University, Canada Stefan Pape, Nipissing
More informationWorksheet 2.7: Critical Points, Local Extrema, and the Second Derivative Test
Boise State Math 275 (Ultman) Worksheet 2.7: Critical Points, Local Extrema, and the Second Derivative Test From the Toolbox (what you need from previous classes) Algebra: Solving systems of two equations
More informationFigure 1: A positive crossing
Notes on Link Universality. Rich Schwartz: In his 1991 paper, Ramsey Theorems for Knots, Links, and Spatial Graphs, Seiya Negami proved a beautiful theorem about linearly embedded complete graphs. These
More informationMinimum Bounding Boxes for Regular Cross-Polytopes
Minimum Bounding Boxes for Regular Cross-Polytopes Salman Shahid Michigan State University shahids1@cse.msu.edu Dr. Sakti Pramanik Michigan State University pramanik@cse.msu.edu Dr. Charles B. Owen Michigan
More informationShape 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 informationShape from Textures Phenomenon for 3D Modeling Applications with Big Data Images
Shape from Textures Phenomenon for 3D Modeling Applications with Big Data Images Kalaivani Thangamani, Takashi Okuma, Ryosuke Ichikari and Takeshi Kurata Service Sensing, Assimilation, and Modeling Research
More informationComputational QC Geometry: A tool for Medical Morphometry, Computer Graphics & Vision
Computational QC Geometry: A tool for Medical Morphometry, Computer Graphics & Vision Part II of the sequel of 2 talks. Computation C/QC geometry was presented by Tony F. Chan Ronald Lok Ming Lui Department
More information