Texture by xiangpeng

VIEWS: 24 PAGES: 9

• pg 1
```									Fast color texture
recognition using
chromaticity moments
Pattern Recognition Letters 21 (2000) 837-841

Presented by Waseem Khatri
Existing approaches to texture analysis
   Statistical – Moments , Co-occurrence matrix
   Model Based – Fractal, Stochastic models
   Structural – Microtexture , Macrotexture , Morphology
   Transform – Fourier , Wavelet , Gabor transforms

Limitations
   Computationally Intensive
   Cannot differentiate subtle variation in textures
   Scaling and Rotation
Proposed Method

   CIE xy chromaticity diagram of an image
   2D and 3D moments to characterize a given color
image.
   Classification using distance measure
CIE XYZ Color Space
Chromaticity:
The quality of a color as
determined by its dominant
wavelength

   Chromaticity diagram is a
two dimensional
representation of an image
where each pixel produces a
pair of (x,y) values
   Matlab: rgb2xyz
2D Shape and 2D distribution

2D Trace            2D Distribution
Moments
Definition:
If f(x,y) is piecewise continuous and has non zero values only
in a finite part of the xy-plane, moments of all orders exist and
the moment sequence (mpq) is uniquely determined by f(x,y).
   
m pq      
  
x p y q f ( x, y ) dxdy

Why moments ?
Moments uniquely capture the nature of both the 2D shape and
the 2D distribution of chromaticities.
Procedure
   Given image is converted into CIE xyz color space
   The trace of the chromaticity diagram is computed
T(x,y) = 1    if exists (i,j) : I(i,j) produces (x,y)
0    otherwise;
0<i<Lx , 0<i<Ly

   The 2D distribution is computed using:
D(x,y)= k , where k= #pixels producing (x,y)

   Moments are computed using:
X s 1Ys 1
M T (m, l )       
x 0 y 0
x m y l T ( x, y)

X s 1Ys 1
M D (m, l )       x
x 0 y 0
m
y l D( x, y)
Classification

   Moments for all the classes in the database are
computed
   Moments for the test sample is computed
   Minimum Distance measure
d=|x-x | where x is the feature vector of the class
i

xi is the feature vector of the test image

   The given test sample is assigned to the class
from which it has the minimum distance
Conclusion
 Simple

 Efficient

 Effective for a database with distinct
texture
 Uses small number of chromaticity
moment features

Drawbacks
 Error rate is high if the database contains
textures that are not significantly different

```
To top