1Ellen L. Walker Matching Find a smaller image in a larger image Applications Find object / pattern of interest in a larger picture Identify moving objects.

Slides:



Advertisements
Similar presentations
Medical Image Registration Kumar Rajamani. Registration Spatial transform that maps points from one image to corresponding points in another image.
Advertisements

電腦視覺 Computer and Robot Vision I
November 12, 2013Computer Vision Lecture 12: Texture 1Signature Another popular method of representing shape is called the signature. In order to compute.
Computer Vision Lecture 18: Object Recognition II
Computer vision: models, learning and inference Chapter 13 Image preprocessing and feature extraction.
Face Alignment with Part-Based Modeling
Chapter 8 Content-Based Image Retrieval. Query By Keyword: Some textual attributes (keywords) should be maintained for each image. The image can be indexed.
Computer vision: models, learning and inference
Announcements Final Exam May 13th, 8 am (not my idea).
University of Pennsylvania 1 GRASP CIS 580 Machine Perception Fall 2004 Jianbo Shi Object recognition.
Each pixel is 0 or 1, background or foreground Image processing to
1Ellen L. Walker Edges Humans easily understand “line drawings” as pictures.
Instructor: Mircea Nicolescu Lecture 13 CS 485 / 685 Computer Vision.
10/11/2001Random walks and spectral segmentation1 CSE 291 Fall 2001 Marina Meila and Jianbo Shi: Learning Segmentation by Random Walks/A Random Walks View.
Announcements Final Exam May 16 th, 8 am (not my idea). Practice quiz handout 5/8. Review session: think about good times. PS5: For challenge problems,
Iterative closest point algorithms
(1) Feature-point matching by D.J.Duff for CompVis Online: Feature Point Matching Detection,
Detecting Image Region Duplication Using SIFT Features March 16, ICASSP 2010 Dallas, TX Xunyu Pan and Siwei Lyu Computer Science Department University.
Lectures 10&11: Representation and description
Automatic Image Alignment (feature-based) : Computational Photography Alexei Efros, CMU, Fall 2006 with a lot of slides stolen from Steve Seitz and.
Circular Augmented Rotational Trajectory (CART) Shape Recognition & Curvature Estimation Presentation for 3IA 2007 Russel Ahmed Apu & Dr. Marina Gavrilova.
CS 450: Computer Graphics REVIEW: OVERVIEW OF POLYGONS
Summarized by Soo-Jin Kim
05 - Feature Detection Overview Feature Detection –Intensity Extrema –Blob Detection –Corner Detection Feature Descriptors Feature Matching Conclusion.
Orthogonal moments Motivation for using OG moments Stable calculation by recurrent relations Easier and stable image reconstruction - set of orthogonal.
Image Formation. Input - Digital Images Intensity Images – encoding of light intensity Range Images – encoding of shape and distance They are both a 2-D.
October 14, 2014Computer Vision Lecture 11: Image Segmentation I 1Contours How should we represent contours? A good contour representation should meet.
1 Interest Operators Harris Corner Detector: the first and most basic interest operator Kadir Entropy Detector and its use in object recognition SIFT interest.
Analysis of shape Biomedical Image processing course, Yevhen Hlushchuk and Jukka Parviainen.
Digital Image Processing Lecture 20: Representation & Description
Local invariant features Cordelia Schmid INRIA, Grenoble.
Segmentation Course web page: vision.cis.udel.edu/~cv May 7, 2003  Lecture 31.
Digital Image Processing Lecture 7: Geometric Transformation March 16, 2005 Prof. Charlene Tsai.
Digital Image Processing CSC331
September 23, 2014Computer Vision Lecture 5: Binary Image Processing 1 Binary Images Binary images are grayscale images with only two possible levels of.
1Ellen L. Walker Object Recognition Instance Recognition Known, rigid object Only variation is from relative position & orientation (and camera parameters)
Digital Image Processing CCS331 Relationships of Pixel 1.
Generalized Hough Transform
Data Extraction using Image Similarity CIS 601 Image Processing Ajay Kumar Yadav.
Digital Image Processing, 2nd ed. © 2002 R. C. Gonzalez & R. E. Woods Representation & Description.
CS654: Digital Image Analysis Lecture 25: Hough Transform Slide credits: Guillermo Sapiro, Mubarak Shah, Derek Hoiem.
A survey of different shape analysis techniques 1 A Survey of Different Shape Analysis Techniques -- Huang Nan.
Pre-AP Unit 7 Vocabulary. Area – the measurement attribute that describes the number of square units a figure or region covers Circumference – a linear.
Local invariant features Cordelia Schmid INRIA, Grenoble.
CS654: Digital Image Analysis Lecture 36: Feature Extraction and Analysis.
Content-Based Image Retrieval QBIC Homepage The State Hermitage Museum db2www/qbicSearch.mac/qbic?selLang=English.
1Ellen L. Walker Matching Find a smaller image in a larger image Applications Find object / pattern of interest in a larger picture Identify moving objects.
October 16, 2014Computer Vision Lecture 12: Image Segmentation II 1 Hough Transform The Hough transform is a very general technique for feature detection.
1Ellen L. Walker 3D Vision Why? The world is 3D Not all useful information is readily available in 2D Why so hard? “Inverse problem”: one image = many.
1 Overview representing region in 2 ways in terms of its external characteristics (its boundary)  focus on shape characteristics in terms of its internal.
776 Computer Vision Jan-Michael Frahm Spring 2012.
Machine Vision Edge Detection Techniques ENT 273 Lecture 6 Hema C.R.
: Chapter 5: Image Filtering 1 Montri Karnjanadecha ac.th/~montri Image Processing.
Image features and properties. Image content representation The simplest representation of an image pattern is to list image pixels, one after the other.
Instructor: Mircea Nicolescu Lecture 10 CS 485 / 685 Computer Vision.
Morphological Image Processing
776 Computer Vision Jan-Michael Frahm Spring 2012.
April 21, 2016Introduction to Artificial Intelligence Lecture 22: Computer Vision II 1 Canny Edge Detector The Canny edge detector is a good approximation.
CSE 554 Lecture 8: Alignment
Image Representation and Description – Representation Schemes
Digital Image Processing Lecture 20: Representation & Description
LOCUS: Learning Object Classes with Unsupervised Segmentation
Feature description and matching
Computer Vision Lecture 5: Binary Image Processing
Fitting Curve Models to Edges
Computer Vision Lecture 16: Texture II
CS654: Digital Image Analysis
Computer and Robot Vision I
Comparing Images Using Hausdorff Distance
Fourier Transform of Boundaries
Presentation transcript:

1Ellen L. Walker Matching Find a smaller image in a larger image Applications Find object / pattern of interest in a larger picture Identify moving objects between a pair of pictures Pre-process for multi-image recognition, e.g. ster You can’t just subtract! Noise Occlusion distortion

2Ellen L. Walker A simple matching algorithm For each position / location / orientation of the pattern Apply matching criteria Local maxima of the criteria exceeding a preset threshold are considered matches (Sound familiar? Replace “match criteria” with “edge strength…”)

3Ellen L. Walker Match Criteria: Digital Difference Take absolute value of pixel differences across the mask Convolution with mask of all -1’s followed by absolute value Not very robust to noise & slight variations Best is 0, so algorithms would minimize rather than maximize Very dark or very light masks will give larger mismatch values!

4Ellen L. Walker Match Criteria (Correlations) All are inverted (and 1 is added to avoid divide by 0) C1 - use largest distance in mask C2 - use sum of distances in mask C3 - use sum of squared distances in mask

5Ellen L. Walker Matching Strategies Brute force (try all possibilities) Hierarchical (coarse-to-fine) Hierarchical (sub-patterns; graph match) Optimized operation sequence Short-circuit processing at mismatches for efficiency

6Ellen L. Walker Matching Higher-level features Elementary geometric properties (boundaries) Boundary length (perimeter) Curvature: perimeter/ count of “direction change” pixels Chord distribution (lengths & angles of all chords) Complete boundary representations Polygonal approximation (recursive line-split algorithm) Contour partitioning (curvature primal sketch) B-Splines

7Ellen L. Walker When do they match? Exact (numerical) match Match within (numerical) tolerance, i.e. threshold numeric difference Match within (geometric) tolerance, i.e. threshold geometric distance to boundary More complex descriptions are harder to match; generally more sensitive to noise!

8Ellen L. Walker Geometric Invariants Given: known shape and known transformation Use: measure that is invariant over the transformation The value is measurable and constant over all xformed shapes Examples Euclidean distance: invariant under translation & rotation Angle between line segments: translation, rotation, scale Cross-ratio: projective invariants (including perspective)

9Ellen L. Walker Geometric Transformations In general, a geometric transformation is any operation on points that yields points Linear transformations can be represented by matrix multiplication: Result is x’/s’, y’/s’

10Ellen L. Walker Example transformations Translation Set diagonals to 1, right column to new location, all else 0 Rotation Set upper four elements to cos(theta), -sin(theta), sin(theta), cos(theta), lower diagonal to 1, all else 0 Scale Set diagonals to 1 and lower right to scale factor Projective transform Any arbitrary 3x3 matrix!

11Ellen L. Walker Cross Ratio Consider four rays “cut” by two lines I = (A-C)(B-D) / (A-D)(B-C)

12Ellen L. Walker Cross Ratio Examples Two images of one object makes 2 matching cross ratios! Dual of cross ratio: four lines from a point instead of four points on a line Any five non-collinear but coplanar points yield two cross-ratios (from sets of 4 lines)

13Ellen L. Walker Using Invariants for Recognition Measure the invariant in one image (or on the object) Find all possible instances of the invariant (e.g. all sets of 4 collinear points) in the (other) image If any instance of the invariant matches the measured one, then you (might) have found the object Research question: to what extent are invariants useful in noisy images?