Texture What is texture? Easy to recognize, hard to define - - PDF document
Texture What is texture? Easy to recognize, hard to define Deterministic textures (thing-like) Stochastic textures (stuff-like) Tasks Discrimination / Segmentation Classification Texture synthesis
1
– Easy to recognize, hard to define – Deterministic textures (“thing-like”) – Stochastic textures (“stuff-like”)
– Discrimination / Segmentation – Classification – Texture synthesis – Shape from texture – Texture transfer – Video textures
2
Shape from Texture
– An image obeying some statistical properties – Similar structures repeated over and over again – Often has some degree of randomness
3
4
Steerable (i.e., Oriented) Pyramids
5
Texture Synthesis [Efros & Leung, ICCV 99]
– What is ? – Find all the windows in the image that match the neighborhood
– To synthesize x
sample image Generated image
SAMPLE x
6
A Markov random field (MRF)
First-order MRF:
D C X A B X ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ X ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
– a sequence of random variables – is the state of the model at time t – Markov assumption: each state is dependent only on the previous one
– The above is actually a first-order Markov chain – An N’th-order Markov chain:
7
sample image
– An exact neighborhood match might not be present – So we find the best matches using SSD error and randomly choose between them, preferring better matches with higher probability SAMPLE
Generated image
x
– Starting from the initial image, “grow” the texture one pixel at a time
8
Increasing window size
9
aluminum wire reptile skin
Growing garbage Verbatim copying
10
11
"
" "
"
12
– Plagiarize as much of the source image as you can – Then try to cover up the evidence
– Texture blocks are by definition correct samples of texture, so the only problem is connecting them together
Constraint Texture sample
13
and “paint” it onto another object
– This requires separating texture and shape – That’s HARD, but we can cheat – Assume we can capture shape by boundary and rough shading
%&'( %&'(
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34