Minimum Likelihood Image Feature and Scale Detection Kim Steenstrup Pedersen Collaborators: Pieter van Dorst, TUe, The Netherlands Marco Loog, ITU, Denmark.

Slides:



Advertisements
Similar presentations
Photon absorption Local heating Thermal expansion INFN-LENS T2 Braginsky et al., Phys. Lett. A 264, 1 (1999) Cerdonio et al., Phys. Rev. D 63, (2001)
Advertisements

Slide 1 Insert your own content. Slide 2 Insert your own content.
QUALITY CONTROL TOOLS FOR PROCESS IMPROVEMENT
Copyright © 2002 Pearson Education, Inc. Slide 1.
By D. Fisher Geometric Transformations. Reflection, Rotation, or Translation 1.
July 11, 2006 Council of Graduate Schools 1 Ph.D. Completion Project: Using the Baseline Data 2006 CGS Summer Workshop Technical Workshop.
Jörg Drechsler (Institute for Employment Research, Germany) NTTS 2009 Brussels, 20. February 2009 Disclosure Control in Business Data Experiences with.
0 - 0.
ALGEBRAIC EXPRESSIONS
DIVIDING INTEGERS 1. IF THE SIGNS ARE THE SAME THE ANSWER IS POSITIVE 2. IF THE SIGNS ARE DIFFERENT THE ANSWER IS NEGATIVE.
MULTIPLYING MONOMIALS TIMES POLYNOMIALS (DISTRIBUTIVE PROPERTY)
SUBTRACTING INTEGERS 1. CHANGE THE SUBTRACTION SIGN TO ADDITION
MULT. INTEGERS 1. IF THE SIGNS ARE THE SAME THE ANSWER IS POSITIVE 2. IF THE SIGNS ARE DIFFERENT THE ANSWER IS NEGATIVE.
Addition Facts
Developing Event Reconstruction for CTA R D Parsons (Univ. of Leeds) J Hinton (Univ. of Leicester)
1 Hierarchical Part-Based Human Body Pose Estimation * Ramanan Navaratnam * Arasanathan Thayananthan Prof. Phil Torr * Prof. Roberto Cipolla * University.
Image Segmentation with Level Sets Group reading
© University of Reading 2007www.reading.ac.uk Sting Jets in severe Northern European Windstoms Suzanne Gray, Oscar Martinez-Alvarado, Laura Baker (Univ.
SADC Course in Statistics Session 4 & 5 Producing Good Tables.
Bayesian network for gene regulatory network construction
Truncation Errors and Taylor Series
Experimental and Quasiexperimental Designs Chapter 10 Copyright © 2009 Elsevier Canada, a division of Reed Elsevier Canada, Ltd.
A Minimum Cost Path Search Algorithm Through Tile Obstacles Zhaoyun Xing and Russell Kao Sun Microsystems Laboratories.
Vanishing Point Detection and Tracking
Université du Québec École de technologie supérieure Face Recognition in Video Using What- and-Where Fusion Neural Network Mamoudou Barry and Eric Granger.
ABC Technology Project
Ter Haar Romeny, FEV Nuclei of fungus cell Paramecium Caudatum Spatial gradient Illumination spectrum -invariant gradient Color RGB original Color-Scale.
Ter Haar Romeny, FEV Geometry-driven diffusion: nonlinear scale-space – adaptive scale-space.
Ter Haar Romeny, EMBS Berder 2004 How can we find a dense optic flow field from a motion sequence in 2D and 3D? Many approaches are taken: - gradient based.
© Charles van Marrewijk, An Introduction to Geographical Economics Brakman, Garretsen, and Van Marrewijk.
© Charles van Marrewijk, An Introduction to Geographical Economics Brakman, Garretsen, and Van Marrewijk.
Image Registration  Mapping of Evolution. Registration Goals Assume the correspondences are known Find such f() and g() such that the images are best.
Insert Date HereSlide 1 Using Derivative and Integral Information in the Statistical Analysis of Computer Models Gemma Stephenson March 2007.
Past Tense Probe. Past Tense Probe Past Tense Probe – Practice 1.
CSCE 643 Computer Vision: Template Matching, Image Pyramids and Denoising Jinxiang Chai.
Addition 1’s to 20.
Probabilistic Tracking and Recognition of Non-rigid Hand Motion
Test B, 100 Subtraction Facts
11 = This is the fact family. You say: 8+3=11 and 3+8=11
Week 1.
Group Meeting Presented by Wyman 10/14/2006
CSE 473/573 Computer Vision and Image Processing (CVIP) Ifeoma Nwogu Lecture 27 – Overview of probability concepts 1.
1/22 Worst and Best-Case Coverage in Sensor Networks Seapahn Meguerdichian, Farinaz Koushanfar, Miodrag Potkonjak, and Mani Srivastava IEEE TRANSACTIONS.
The STARTS Model David A. Kenny December 15, 2013.
Ter Haar Romeny, Computer Vision 2014 Geometry-driven diffusion: nonlinear scale-space – adaptive scale-space.
Space-time interest points Computational Vision and Active Perception Laboratory (CVAP) Dept of Numerical Analysis and Computer Science KTH (Royal Institute.
Uncertainty Representation. Gaussian Distribution variance Standard deviation.
Lecture 6: Feature matching CS4670: Computer Vision Noah Snavely.
Midterm Review CS485/685 Computer Vision Prof. Bebis.
EE663 Image Processing Edge Detection 2 Dr. Samir H. Abdul-Jauwad Electrical Engineering Department King Fahd University of Petroleum & Minerals.
Lecture 4: Edge Based Vision Dr Carole Twining Thursday 18th March 2:00pm – 2:50pm.
CS4670: Computer Vision Kavita Bala Lecture 8: Scale invariance.
EE462 MLCV 1 Lecture 3-4 Clustering (1hr) Gaussian Mixture and EM (1hr) Tae-Kyun Kim.
Overview Introduction to local features
Overview Harris interest points Comparing interest points (SSD, ZNCC, SIFT) Scale & affine invariant interest points Evaluation and comparison of different.
Learning the Appearance and Motion of People in Video Hedvig Sidenbladh, KTH Michael Black, Brown University.
Lecture 7: Features Part 2 CS4670/5670: Computer Vision Noah Snavely.
Overview Introduction to local features Harris interest points + SSD, ZNCC, SIFT Scale & affine invariant interest point detectors Evaluation and comparison.
Edge Segmentation in Computer Images CSE350/ Sep 03.
Learning Image Statistics for Bayesian Tracking Hedvig Sidenbladh KTH, Sweden Michael Black Brown University, RI, USA
Image Filtering Spatial filtering
- photometric aspects of image formation gray level images
Corners and Interest Points
Edge detection Goal: Identify sudden changes (discontinuities) in an image Intuitively, most semantic and shape information from the image can be encoded.
Lecture 2: Edge detection
Filtering Things to take away from this lecture An image as a function
Lecture 5: Feature invariance
Filtering An image as a function Digital vs. continuous images
Lecture 5: Feature invariance
Review and Importance CS 111.
Presentation transcript:

Minimum Likelihood Image Feature and Scale Detection Kim Steenstrup Pedersen Collaborators: Pieter van Dorst, TUe, The Netherlands Marco Loog, ITU, Denmark

Gaussian Processes in Practice 2 What is an image feature? Marr’s (1982) primal sketch (edges, bars, corners, blobs) Marr’s (1982) primal sketch (edges, bars, corners, blobs) Geometrical features, Marr’s features defined by differential geometry: Canny (1986), Lindeberg (1998) Geometrical features, Marr’s features defined by differential geometry: Canny (1986), Lindeberg (1998) Iconic features: Koenderink (1993), Griffin & Lillholm (2005) Iconic features: Koenderink (1993), Griffin & Lillholm (2005) Observation: Features are usually points and curves, i.e. sparsely distributed in space (unlikely events). Features have an intrinsic scale / size. How blurred is the edge? What is the size if a bar?

Gaussian Processes in Practice 3 A probabilistic primal sketch Our definition: Features are points that are unlikely to occure under an image model. Similarly the scale of the feature is defined as the most unlikely scale. We use fractional Brownian images as a generic model of the intensity correlation found in natural images. Captures second order statistics of generic image points (non-feature points). We use fractional Brownian images as a generic model of the intensity correlation found in natural images. Captures second order statistics of generic image points (non-feature points). The model includes feature scale naturally. The model includes feature scale naturally. This leads to a probabilistic feature and scale detection. This leads to a probabilistic feature and scale detection. Possible applications: Feature detection, interest points for object recognition, correspondance in stereo, tracking, etc.

Gaussian Processes in Practice 4 Probabilistic feature detection Feature detection: Konishi et al. (1999, 2002, 2003) Konishi et al. (1999, 2002, 2003) Lillholm & Pedersen (2004) Lillholm & Pedersen (2004) Scale selection: Pedersen & Nielsen (1999) Pedersen & Nielsen (1999) Loog et al. (2005) Loog et al. (2005)

Gaussian Processes in Practice 5 Linear scale-space derivatives Scale-space derivatives:

Gaussian Processes in Practice 6 Scale Space k-Jet Representation We use the k-jet as representation of the local geometry: (The coefficients of the truncated Taylor expansion of the blurred image.) Biologically plausible representation (Koenderink et al., 1987)

Gaussian Processes in Practice 7 Probabilistic image models Key results on natural image statistics: Scale invariance / Self-similarity: Power spectrum, : Field (1987), Ruderman & Bialek (1994) Scale invariance / Self-similarity: Power spectrum, : Field (1987), Ruderman & Bialek (1994) In general non-Gaussian filter responses! In general non-Gaussian filter responses! Fractional Brownian images as model of natural images: Mumford & Gidas (2001), Pedersen (2003), Markussen et al. (2005) Mumford & Gidas (2001), Pedersen (2003), Markussen et al. (2005) Jet covariance of natural images resembles that of fractional Brownian images: Pedersen (2003) Jet covariance of natural images resembles that of fractional Brownian images: Pedersen (2003)

Gaussian Processes in Practice 8 Fractional Brownian images

Gaussian Processes in Practice 9 FBm in Jet space (Result from Pedersen (2003))

Gaussian Processes in Practice 10 Detecting Features and Scales Detecting points in scale-space that are locally unlikely (minima): (We could also have maximised.)

Gaussian Processes in Practice 11 Why minimum likeli scales? Lindeberg (1998) maximises polynomials of derivatives in order to detect features and scales. Similarly, we maximise in order to detect features and scales. The difference lies in the choice of polynomial! We use an image model and Lindeberg uses a feature model.

Gaussian Processes in Practice 12 Synthetic examples: Double blobs

Gaussian Processes in Practice 13 Synthetic examples: Blurred step edge

Gaussian Processes in Practice 14 Real Example: Sunflowers

Gaussian Processes in Practice 15 Sunflowers: Multi-scale

Gaussian Processes in Practice 16 Sunflowers: Fixed scale

Gaussian Processes in Practice 17 Summary Minimising the likelihood of an image point under the fractional Brownian image model detects feature points and their intrinsic scale. Minimising the likelihood of an image point under the fractional Brownian image model detects feature points and their intrinsic scale. There is a relationship between feature types and the  parameter. There is a relationship between feature types and the  parameter. Why over estimation of the scale? Why over estimation of the scale? Preliminary results look promising, a performance evaluation is needed (task based?). Preliminary results look promising, a performance evaluation is needed (task based?). The method is pointwise. How to handle curve features (edges, bars, ridges)? The method is pointwise. How to handle curve features (edges, bars, ridges)?