CSE527-l7 - Feature Extraction Features: local meaningful...

Info iconThis preview shows pages 1–10. Sign up to view the full content.

View Full Document Right Arrow Icon
1 Feature Extraction • Features: local meaningful detectable • Points • Edges – Step edges – Line edges – Contours – Deformable models • Closed contours are boundaries • Regions 2 Gabor filterbanks: consider combinations of and Spatially localized frequency analysis Increasing Previous Lecture: Gabor filters Horizontal Frequency Vertical Frequency Fourier transform isocurves Increasing
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
2 Envelope estimation for a 2D sinusoidal 3 Convolve Textons (Malik et al, IJCV 2001) Filterbank Image Database
Background image of page 2
3 ….. frequency codewords Object recognition as texture classification Pattern recognition task
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
4 Template Matching • The most straightforward approach to finding a particular pattern in the image is to look at every patch of pixels and to directly compare the intensity values in that patch to the intensity values in the search pattern. • Consider two vectors of intensity values, x and y, One way to compare them is by simply computing the Sum of Squared Differences (SSD) Line Fitting • Follows edge extraction and linking • Used to find straight line features in an image • Simplest way split and merge techniques – Problem with corners
Background image of page 4
5 Parameterizing lines • For our purposes lines in the image will be parameterized by a vector of 3 numbers (m x , m y , m z ) where: m x 2 + m y 2 = 1 • Points on the line are defined by the equation: m x *x + m y *y + m z = 0 m z (m x , m y ) Least Squares Fitting • Given a set of points (xi, yi) the goal of the fitting procedure is to find the parameters (m x , m y , m z ) which represent the best line
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
6 Least squares criterion • The best fit is defined in terms of the parameters which minimize the sum of the squared residuals Steps • Compute the centroid of the point set • Change coordinates such that the new centroid is (0,0)
Background image of page 6
7 Steps (II) • Solve for (m x , m y ) by minimizing the following quadratic form • with e m m m x m y x y x i y i i ( , ) ( ) = + 2 Steps (III) • Solve for m z
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
8 Issues • The main problem with least squares fitting techniques is that they are heavily influenced by outliers • Solutions to this problem include – Robust weighting measures – Iteratively reweighted least squares – Least median squares The Hough Transform • Another approach that can be used to detect features such as lines in an image is the Hough transform • This approach is a voting scheme based on accumulating evidence in a parameter space
Background image of page 8
9 The Hough Transform
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 10
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 34

CSE527-l7 - Feature Extraction Features: local meaningful...

This preview shows document pages 1 - 10. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online