You are on page 1of 21

3D Video Processing: Utilizing MATLAB

Prof. Kang-Sun Choi School of Electrical, Electronics, and Communication Engineering

http://cv3vpl.koreatech.ac.kr ks.choi@koreatech.ac.kr

12 4 23

Scale Invariant Feature Transform (SIFT)


http://www.vlfeat.org/~vedaldi/code/sift.html

2
12 4 23

MEX Compiling
MATLAB is an interpreter For higher performance Implemented by C & Compiled by C compiler Code optimization for each system

3
12 4 23

Example of .c le
include mex.h mexFunction() mex(imsmooth.c, opt);

4
12 4 23

Example of .c le
include mex.h mexFunction() mex(imsmooth.c, opt);

4
12 4 23

Example of .c le
include mex.h mexFunction() mex(imsmooth.c, opt);

4
12 4 23

Notes
Use mex les like general MATLAB functions A = imsmooth( I, sigma ); imread( ) returns uint8 data. [0~255] need to convert double type & optionally normalize it help / doc size ( ) reshape ( ) A = [ ]; imread ( ) imshow ( ) gure ( ) gradient ( )
5
12 4 23

Image Gradient

6
12 4 23

Image Gradient

I =
I x
12 4 23

I x

I y

= I[x, y] I[x 1, y]

I y

= I[x, y] I[x, y 1]

Image Gradient

r[x,y] 204 g[x,y] 205 b[x,y] 209

204 207 217

178 176 170

36 37 32

0 0 16

I =
I x
12 4 23

I x

I y

= I[x, y] I[x 1, y]

I y

= I[x, y] I[x, y 1]

Image Gradient
L1 Norm of Gradient
I x

I y

8
12 4 23

Image Gradient

9
12 4 23

Hessian Matrix
A=

I I I I I I
2 x x y 2 y

x y

10
12 4 23

Difference of Gaussians (DoG)

11
12 4 23

Difference of Gaussians (DoG)

12
12 4 23

Difference of Gaussians (DoG)

12
12 4 23

Difference of Gaussians (DoG)

13
12 4 23

Gaussian Pyramid

14
12 4 23

Gaussian Pyramid

15
12 4 23

Gaussian Pyramid

16
12 4 23

Gaussian Pyramid

16
12 4 23

You might also like