Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal...

17
Texture 1

Transcript of Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal...

Page 1: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Texture

1

Page 2: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Today

  Key Question: How do we represent texture?

  Topics   Definition of texture   Texture segmentation   Texture analysis   Texture synthesis   Shape from texture (only statement of

the problem)

2

Page 3: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

What is texture?

  No formal definition exists   Sonka, Hlavac and Boyle:

  “something consisting of mutually related elements”

  Trucco and Verri:   “A surface texture is created by the regular repetition of an

element or pattern, called surface textel, on a surface”   “An image texture is the image of a surface texture, itself

a repetition of image texels, the shape of which is distorted by the projection across the image”

3

Page 4: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Texture segmentation

  Question: Is texture the property of a point or of a region?

  We need a region to have a texture!   This is a “chicken and egg” problem.

  Texture segmentation can be done can be done by detecting boundaries of a region characterized by similar texture

  Texture boundaries can be detected using standard edge detection techniques (applied to the texture measures determined at each point)

  We typically use a local window to estimate texture properties and assign those texture properties as point properties of the windows’ center row and column

4

Page 5: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Texture descriptors

  Measures of smoothness, coarseness, and regularity

5

Page 6: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Approaches for texture description

  Statistical:   Describe texture as smooth, coarse, grainy etc.   Scale dependent!

  Structural:   Deal with the arrangement of image primitives. Example:

regularly spaced parallel lines   Tone and structure of a texture

  Tone=based on pixel intensity properties in a primitive   Structure: the spatial relationship between the primitives

  Spectral techniques   Good for analyzing periodic or quasi-periodic textures

6

Page 7: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Statistical approaches

  Use the statistical moments of the intensity histogram of an image or region   Order 1: Mean   Order 2: Variance

 Normalized smoothness descriptor   Order 3: Skewness (symmetry of the

histogram)   Order 4: Kurtosis (flatness of the histogram)   Additional measures: uniformity, enthropy

7

Page 8: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

8

Page 9: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Limits of statistical moments of intensity histogram

  Carry no information about the relative position of pixels with respect to each other

  Don’t tell us anything about ‘texels’

9

Page 10: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Co-occurrence matrice

10

Page 11: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”
Page 12: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Spectral approaches

  Excellent for detecting periodic textures

12

Page 13: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”
Page 14: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Texture synthesis

  Alexei Efros and Thomas Leung: Texture synthesis by non-parametric sampling, ICCV 1999 (mandatory reading)

  http://graphics.cs.cmu.edu/people/efros/research/NPS/efros-iccv99.ppt

14

Page 15: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Goal  of  Texture  Synthesis  

•  Given  a  finite  sample  of  some  texture,  the  goal  is  to  synthesize  other  samples  from  that  same  texture.    –  The  sample  needs  to  be  "large  enough"  

True (infinite) texture

SYNTHESIS

generated image

input image

Page 16: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

The  Challenge  

•  Texture  analysis:  how  to  capture  the  essence  of  texture?    

•  Need  to  model  the  whole  spectrum:  from  repeated  to  stochasDc  texture  

•  This  problem  is  at  intersecDon  of  vision,  graphics,  staDsDcs,  and  image  compression  

repeated

stochastic

Both?

Page 17: Texture - UVic.caaalbu/computer vision 2010/L26. Texture.pdf · What is texture? No formal definition exists Sonka, Hlavac and Boyle: “something consisting of mutually related elements”

Shape  from  texture  

17

“An image texture is the image of a surface texture, itself a repetition of image texels, the shape of which is distorted by the projection across the image”