lec6-prelim

lec6-prelim - Lecture 6 HW1 due today Last Class...

Info iconThis preview shows pages 1–7. Sign up to view the full content.

View Full Document Right Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 1 Lecture 6: Sept 13, 10 HW1 due today Last Class Segmentation methods: histogram based, k-means clustering Intro to mean shift filtering Today’s objective Tutorial on OpenCV Mean shift segmentation Normalized cuts
Background image of page 1

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

View Full DocumentRight Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 2 Mean Shift Filtering From work of Comaniciu and Meer (two papers posted on the class web page); also see RS, Section 5.3.2 Filtering while preserving regions/edges General idea is to estimate the maxima of the probability density function given only some sample points (drawn according to the density function) Where are the clusters figure (b), RS fig 5.16
Background image of page 2
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 3 Kernel Density Estimation (RS 5.3.2) f(x) is the estimate, K(x) is the kernel function, G(x) is derivative of K(x) Direct computation can be expensive, instead, compute maxima using the mean-shift method
Background image of page 3

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

View Full DocumentRight Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 4 Mean Shift Filtering General idea Choose a neighborhood Average over points in the neighborhood (may be weighted by a Gaussian or some other kernel function) to compute a mean Shift center of neighborhood to new mean (hence mean shift ) Repeat until convergence (guaranteed with proper choice of kernel and shift steps) Replace range of point with that of convergent point Can be shown that the procedure results in estimating local maximum of f(x) Consider all points converging to the same maximum to correspond to the same cluster, assign them the value of the maximum (for filtering)
Background image of page 4
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 5 Example Binary function in 2-D
Background image of page 5

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

View Full DocumentRight Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia
Background image of page 6
Image of page 7
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 26

lec6-prelim - Lecture 6 HW1 due today Last Class...

This preview shows document pages 1 - 7. Sign up to view the full document.

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