Lecture-8-h - Lecture-8 Haralick’s Edge Detector...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Lecture-8 Haralick’s Edge Detector Haralick’s Edge Detector • Fit a bi-quadratic polynomial to a small neighborhood of a pixel. • Compute analytically second and third directional derivatives in the direction of gradient. • If the second derivative is equal to zero, and the third derivative is negative, then that point is an edge point. 1 Haralick’s Edge Detector Bi-cubic polynomial: Gradient angle, defined with positive y-axis: Homework Directional derivative Gradient angle, defined with positive x-axis: Haralick’s Edge Detector 2 Haralick’s Edge Detector Homework Haralick’s Edge Detector 3 Haralick’s Edge Detector First order polynomial 9 points give 9 eqs Haralick’s Edge Detector 4 Computing coefficients using convolution 5 Haralick’s Edge Detector Comparison of Three Edge Detectors • Marr-Hildreth – Gaussian filter – Zerocrossings in Laplacian • Canny – Gaussian filter – Maxima in gradient magnitude • Haralick – Smoothing through bi-cubic polynomial – Zerocrossings in the second directional derivative, and negative third derivative 6 Laplacian and the second Directional Derivative and the direction of Gradient D2 f = f xx + f yy = fq'' + f n'' fq' = f x cos q + f y sin q fq'' = ( f xx cosq + f yx sin q ) cosq + ( f xy cosq + f yy sin q ) sin q fq'' = f xx cos2 q + f yy sin 2 q + 2 f xy cosq sin q † f n'' = f xx cos2 n + f yy sin 2 n + 2 f xy cos n sin n † '' f n = f xx cos2 (q + 90) + f yy sin2 (q + 90) + 2 f xy cos(q + 90) sin(q + 90) † f n'' = f xx sin 2 q + f yy cos2 q - 2 f xy cosq sin q † † † Laplacian and the second Directional Derivative and the direction of Gradient fq'' = f xx cos2 q + f yy sin 2 q + 2 f xy cosq sin q f n'' = f xx sin 2 q + f yy cos2 q - 2 f xy cosq sin q † D2 f = f xx + f yy = fq'' + f n'' † 7 Scales • What should be sigma value for Canny and LG edge detection? – Marr-Hildreth: • If use multiple sigma values (scales), how do you combine multiple edge maps? – Spatial Coincidence assumption: • Zerocrossings that coincide over several scales are physically significant. Scale Space • Apply whole spectrum of scales • Plot zerocrossings vs scales in a scale-space • Interpret scale space contours – Contours are arches, open at the bottom, closed at the top – Interval tree • Each interval I corresponds to a node in a tree, whose parent node represents larger interval, from which interval I emerged, and whose off springs represents smaller intervals into which I subdivides. • Stability of a node is a scale range over which the interval exits. 8 Scale Space • Top level description – Iteratively remove nodes from the tree, splicing out nodes that are less stable than any of their parents and off springs Scale Space Multiple smooth versions of a signal Zerocrossings at multiple scale 9 Scale Space Scale Space Interval Tree Scale Space A top level description of several signals using stability criterion. 10 ...
View Full Document

This note was uploaded on 06/12/2011 for the course CAP 5415 taught by Professor Staff during the Fall '08 term at University of Central Florida.

Ask a homework question - tutors are online