feature detection

9
Computer Vision September 2002 L1.1 © 2002 by Davi Geiger Feature Detection Feature Detection

Upload: qamra

Post on 05-Jan-2016

33 views

Category:

Documents


0 download

DESCRIPTION

Feature Detection. Feature Detection. Image Features –Decisions!. Features such as edges, corners, junction, eyes, … are obtained by making some decision from the image measurements. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Feature Detection

Computer Vision September 2002 L1.1© 2002 by Davi Geiger

Feature Detection

Feature Detection

Page 2: Feature Detection

Computer Vision September 2002 L1.2© 2002 by Davi Geiger

Image Features –Decisions!

Features such as edges, corners, junction, eyes, … are obtained by making some decision from the image measurements.

Decisions are the result of some comparison followed by a choice. Examples (i) if a measurement is above a threshold we accept, not otherwise; (ii) if a measurement is the largest compared to others, we select it.

Page 3: Feature Detection

Computer Vision September 2002 L1.3© 2002 by Davi Geiger

Decisions: Edgels (Edge-pixels and Orientation)

)|),,,(ˆ||),2/,,(ˆ|(maxarg),(max syxIDsyxIDyxs

TsyxIDsyxID )|),,,(ˆ||),2

,,(ˆ|( such that Edge threshold: Decision!

Edge orientation:Decision!

A step edge at The value is (equally) large for both and as shown (in red) for scale s=3 pixels. It is also large for values not shownHowever, the quantity is significantly larger for

|),2/,,(~

| syxID

|),,,(~

||),2/,,(~

| syxIDsyxID

                               

                               

                               

                               

                           

                               

                               

                             

                               

),,,(ˆ syxID ),

2,,(ˆ syxID

3,

4,

6,0

Page 4: Feature Detection

Computer Vision September 2002 L1.4© 2002 by Davi Geiger

),(3max yx

The gray level indicates the angle: the darkest one is 0 degrees. The larger is the angle the lighter is displayed, up to

|),2/,,(ˆ|

||),,,(ˆ||),2/,,(ˆ||),(

max

maxmaxmax

syxID

syxIDsyxIDyx

s

sss

Strength of the Edgel

Decisions: Edgels (cont.)

HTsyxH ),,(ˆ

Eliminate some spurious locations. Decision!

Page 5: Feature Detection

Computer Vision September 2002 L1.5© 2002 by Davi Geiger

Decisions: Local Angle Change

),()sin,cos(),( maxmaxmax yxyyxxyx sss Angle change

y

x

),(max yxs

2)sin,cos(max

yyxxs

2),(max

yxs

A contour segment

),( yxs

212222 )sin,cos( yx

is the contour curvature multiplied by the arc length , where

),(max yxs .),(max yxs where

Page 6: Feature Detection

Computer Vision September 2002 L1.6© 2002 by Davi Geiger

TsyxID |),,,(~

|such that

HTsyxH ),,(~

Decisions: Junctions, Corners

Junction threshold: Decision!

Eliminate spurious locations. Decisions!

Examining the values of where allow us to characterize the junctions. For example, when only two value of pass the test and or suggest a corner. Corners are many times called L-junctions. If three angles are detected, it may be a T-junction or an Y-junction. T-junctions exhibit one region with angle near and usually arise in images due to surface occlusions in a scene. Four angles suggest a X-junction, and usually arise in images due to surface transparency in the scene.

Note that this detector also detects many edgels.

65),(),( 122 yxyx

Remove (Undo) detection if

65)2( 2

TsyxID |),,,(~

|

Page 7: Feature Detection

Computer Vision September 2002 L1.7© 2002 by Davi Geiger

Decisions: Connecting Edgels, Pseudocode

Algorithm to link edgels. Start with a seed location (xc,yc)

Contour-Follower(xc , yc)

if (Edgel(xc , yc ) NIL )

Link-neighbors+(xc , yc ,max)

Link-neighbors-(xc , yc ,max)

end

Link-neighbors±(xc , yc ,max)

xn± = xc ± xmax cosmax ;

yn± = yc ± ymax sinmax ;

if (Edgel(xn±, yn±) NIL )

Link((xc , yc), (xn±, yn±))

Link-neighbors±( xn± , yn± ,max(xn± , yn±) )

end

max

(xc , yc)

Page 8: Feature Detection

Computer Vision September 2002 L1.8© 2002 by Davi Geiger

Decisions: Connecting Edgels, Pseudocode

Link-neighbors±(xc , yc ,max-c)

xn± = xc ± xmax cosmax ;

yn± = yc ± ymax sinmax ;

if ((Edgel(xn±, yn±) NIL )&(Coherence(xc , yc xn±, yn±)))

Link((xc , yc), (xn±, yn±))

Link-neighbors±( xn± , yn± ,max-n(xn± , yn±) )

endCoherence(xc , yc xn, yn )

if ( and ) return True else Nilend

1maxmax |),2

,,(~

),2

,,(~

| TsyxIsyxI cccnnn

1maxmax |),2

3,,(

~),

2

3,,(

~| TsyxIsyxI cccnnn

Page 9: Feature Detection

Computer Vision September 2002 L1.9© 2002 by Davi Geiger

Threshold Parameters: Estimation

We have considered at least four parameters: 1,,, TTTT H

How to estimate them? One technique is Histogram partition:

Plot the Histogram and find the parameter that “best partition it”:

|),2/,,(ˆ| syxID

ountingHistogram c:

T