G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

27
Sharif University of Technology A modified algorithm to obtain Translation, Rotation & Scale invariant Zernike Moment shape Descriptors G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

description

Sharif University of Technology A modified algorithm to obtain Translation, Rotation & Scale invariant Zernike Moment shape Descriptors. G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli. Introduction. Shape is one of the most important features to human for visual distinguishing system. - PowerPoint PPT Presentation

Transcript of G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

Page 1: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

Sharif University of Technology

A modified algorithm to obtain Translation, Rotation & Scale

invariant Zernike Moment shape Descriptors

G.R. AmayehDr. S. Kasaei

A.R. Tavakkoli

Page 2: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

2

Introduction Shape is one of the most important

features to human for visual distinguishing system.

Shape Descriptors Contour-Base

Using contour information Neglect image details

Region-Base Using region information

Page 3: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

3

Shape Descriptors

Fig.1: Same regions. Fig.2: Same contours.

Page 4: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

4

Zernike & Pseudo-Zernike Moments

Zernike Moments of Order n, with m-repetition:

Zernike Moment’s Basis Function jm

mnmnmn eRVyxV )(),(),( ,,,

CircleUnit mnmn dydxyxVyxfn

Z ,,1 *

,,

evenismn

nmWhere

(1)

(2)

(3)

Page 5: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

5

Zernike & Pseudo-Zernike Moments

Zernike Moment Radial Polynomials:

Pseudo-Zernike Radial Polynomials:

MZforS

mnS

mnS

snR

mn

s

SnS

mn

2

||

0

2,

)!2

||()!

2

||(!

)!()1(

MZPsforSmnSmnS

snR

mn

s

SnS

mn .)!||()!||(!

)!12()1(||

0,

(4)

(5)

Page 6: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

6

A Cross Section ofRadial Polynomials of ZM & PsZM

Fig.3 : ZM (blue) & Ps. ZM (red) of 4-order with repetition 0.

Fig.5 : ZM (blue) & Ps. ZM (red) of 5-order with repetition 1.

Fig.4 : ZM (blue) & Ps. ZM (red) of 6-order with repetition 4.

Fig.6 : ZM (blue) & Ps. ZM (red) of 7-order with repetition 3.

Page 7: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

7

3-D Illustration of Radial Polynomials of ZM & Ps.ZM

Fig.7 : Radial polynomial of ZM of 7-order with repetition 1.

Fig.8 : Radial polynomial of Ps. ZM of 7-order with repetition 1.

Page 8: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

8

Zernike Moments Properties Invariance Properties:

Zernike Moments are Rotation Invariant Rotation changes only moment’s phase.

Variance Properties: Zernike Moments are Sensitive to

Translation & Scaling.

Page 9: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

9

Achieving Invariant Properties What is needed in segmentation problem?

Moments need to be invariant to rotation, scale and translation.

Solution to achieve invariant properties Normalization method. Improved Zernike Moments without Normalization

(IZM). Proposed Method.

Page 10: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

10

Normalization Method

Algorithm: Translate image’s center of mass to origin.

Scale image:

0,0

1,0

0,0

0,1 ,m

my

m

mxwhere

)6(, yyxxf

0,0

)7(,m

awherea

y

a

xf

Page 11: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

11

Normalization Method

Fig.9 : From left to right, Original, Translated, & Scaled images

Page 12: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

12

Normalization Method

Fig.10 : From left to right, original image & normalized images with different s.

Page 13: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

13

Normalization Method Drawbacks

Interpolation Errors: Down sampling image leads to loss of

data. Up sampling image adds wrong

information to image.

Page 14: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

14

Improved Zernike Moments without Normalization

Algorithm: Translate image’s center of mass to origin. Finding the smallest surrounding circle and

computing ZMs for this circle.

Normalize moments:0,0

,, m

ZZ mn

mn (8)

Fig.11 : Images & fitted circles.

Page 15: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

15

Drawbacks

Increased Quantization Error. Since the SSC of images have a small

number of pixels, image’s resolution is low and this causes more QE.

Page 16: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

16

Proposed Method Algorithm:

Computing a Grid Map. Performing translation and scale on the map

indexes.

Fig.12: Mapping.

Page 17: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

17

Proposed Method Translate origin of coordination system to the center

of mass

(9)

yyy

xxx

Fig(13). Translation of Coordination Origin.

Page 18: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

18

Proposed Method Scale coordination system

yay

xax(10)

Page 19: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

19

Proposed Method Computing Zernike Moment in new

coordinate for where .

We can show that the moments of in the new coordinate system are equal to the moments of in the old coordinate system.

),(),( 2 yxfayxg ),( yx ),( yxg

),( yxg

),(a

y

a

xf

Page 20: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

20

Proposed Method

Fig.15 : From left to right, original image & normalized images with different s.

Page 21: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

21

Proposed Method Special case

Fig.17 : Zernike moments by proposed method & IZM (Improved ZM with out normalization ).

Fig.16 : Original image.

Page 22: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

22

Experimental Results

Fig.16 : Original image & 70% scaled image.

Fig.17 : Error of Zernike moments between original image & scaled image.

Page 23: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

23

Experimental Results

Fig.18 : Original image & 55 degree rotated image.

Fig.19 : Error of Zernike moments between original image & rotated image.

Page 24: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

24

Experimental Results

Fig.21 : Error of Zernike moments between original & scaled images.

Fig.20 : Original image & 120% scaled image.

Page 25: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

25

Experimental Results

Fig.23 : Error of Zernike moments between original image & rotated image.

Fig.21 : Original image & 40 degree rotated image.

Page 26: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

26

Conclusions Principle of our method is same as the

Normalization method. Does not resize the original image.

No Interpolation Error. Reduces the quantization error. (using beta

parameter) Trade off Between QE and power of

distinguishing. Has all the benefits of both pervious methods.

Page 27: G.R. Amayeh Dr. S. Kasaei A.R. Tavakkoli

The End