Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.
-
Upload
horace-heath -
Category
Documents
-
view
217 -
download
1
Transcript of Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.
![Page 1: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/1.jpg)
![Page 2: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/2.jpg)
![Page 3: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/3.jpg)
Thinking in Frequency
Computer Vision
James Hays
Slides: Hoiem, Efros, and others
![Page 4: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/4.jpg)
Office hours• TAs should be available Tuesday and Thursday
2 to 6pm
![Page 5: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/5.jpg)
Recap of Monday
• Linear filtering is dot product at each position– Not a matrix multiplication– Can smooth, sharpen, translate
(among many other uses)
• Be aware of details for filter size, extrapolation, cropping
111
111
111
![Page 6: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/6.jpg)
Review: questions
1. Write down a 3x3 filter that returns a positive value if the average value of the 4-adjacent neighbors is less than the center and a negative value otherwise
2. Write down a filter that will compute the gradient in the x-direction:gradx(y,x) = im(y,x+1)-im(y,x) for each x, y
Slide: Hoiem
![Page 7: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/7.jpg)
Review: questions
3. Fill in the blanks:a) _ = D * B b) A = _ * _c) F = D * _d) _ = D * D
A
B
C
D
E
F
G
H I
Filtering Operator
Slide: Hoiem
![Page 8: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/8.jpg)
Today’s Class
• Fourier transform and frequency domain– Frequency view of filtering– Hybrid images– Sampling
• Reminder: Read your textbook– Today’s lecture covers material in 3.4
Slide: Hoiem
![Page 9: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/9.jpg)
Why does the Gaussian give a nice smooth image, but the square filter give edgy artifacts?
Gaussian Box filter
![Page 10: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/10.jpg)
Hybrid Images
• A. Oliva, A. Torralba, P.G. Schyns, “Hybrid Images,” SIGGRAPH 2006
![Page 11: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/11.jpg)
Why do we get different, distance-dependent interpretations of hybrid images?
?
Slide: Hoiem
![Page 12: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/12.jpg)
![Page 13: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/13.jpg)
![Page 14: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/14.jpg)
Why does a lower resolution image still make sense to us? What do we lose?
Image: http://www.flickr.com/photos/igorms/136916757/ Slide: Hoiem
![Page 15: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/15.jpg)
Thinking in terms of frequency
![Page 16: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/16.jpg)
Jean Baptiste Joseph Fourier (1768-1830)had crazy idea (1807):
Any univariate function can be rewritten as a weighted sum of sines and cosines of different frequencies.
• Don’t believe it? – Neither did Lagrange,
Laplace, Poisson and other big wigs
– Not translated into English until 1878!
• But it’s (mostly) true!– called Fourier Series– there are some subtle
restrictions
...the manner in which the author arrives at these equations is not exempt of difficulties and...his
analysis to integrate them still leaves something to be desired on the score of generality and even
rigour.
Laplace
LagrangeLegendre
![Page 17: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/17.jpg)
How would math have changed if the Slanket or Snuggie had been invented?
![Page 18: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/18.jpg)
A sum of sinesOur building block:
Add enough of them to get any signal g(x) you want!
xAsin(
![Page 19: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/19.jpg)
Frequency Spectra• example : g(t) = sin(2πf t) + (1/3)sin(2π(3f)
t)
= +
Slides: Efros
![Page 20: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/20.jpg)
Frequency Spectra
![Page 21: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/21.jpg)
= +
=
Frequency Spectra
![Page 22: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/22.jpg)
= +
=
Frequency Spectra
![Page 23: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/23.jpg)
= +
=
Frequency Spectra
![Page 24: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/24.jpg)
= +
=
Frequency Spectra
![Page 25: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/25.jpg)
= +
=
Frequency Spectra
![Page 26: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/26.jpg)
= 1
1sin(2 )
k
A ktk
Frequency Spectra
![Page 27: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/27.jpg)
Example: Music• We think of music in terms of frequencies at
different magnitudes
Slide: Hoiem
![Page 28: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/28.jpg)
Other signals• We can also think of all kinds of other signals
the same way
xkcd.com
![Page 29: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/29.jpg)
Fourier analysis in images
Intensity Image
Fourier Image
http://sharp.bu.edu/~slehar/fourier/fourier.html#filtering
![Page 30: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/30.jpg)
Fourier Transform• Fourier transform stores the magnitude and phase at each
frequency– Magnitude encodes how much signal there is at a particular frequency– Phase encodes spatial information (indirectly)– For mathematical convenience, this is often notated in terms of real
and complex numbers
22 )()( IRA )(
)(tan 1
R
IAmplitude: Phase:
![Page 31: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/31.jpg)
Salvador Dali invented Hybrid Images?
Salvador Dali“Gala Contemplating the Mediterranean Sea, which at 30 meters becomes the portrait of Abraham Lincoln”, 1976
![Page 32: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/32.jpg)
![Page 33: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/33.jpg)
![Page 34: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/34.jpg)
Fourier Bases
This change of basis is the Fourier Transform
Teases away fast vs. slow changes in the image.
![Page 35: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/35.jpg)
Fourier Bases
in Matlab, check out: imagesc(log(abs(fftshift(fft2(im)))));
![Page 36: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/36.jpg)
Man-made Scene
![Page 37: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/37.jpg)
Can change spectrum, then reconstruct
![Page 38: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/38.jpg)
Low and High Pass filtering
![Page 39: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/39.jpg)
The Convolution Theorem• The Fourier transform of the convolution of two
functions is the product of their Fourier transforms
• Convolution in spatial domain is equivalent to multiplication in frequency domain!
]F[]F[]F[ hghg
]]F[][F[F* 1 hghg
![Page 40: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/40.jpg)
Filtering in spatial domain-101
-202
-101
* =
![Page 41: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/41.jpg)
Filtering in frequency domain
FFT
FFT
Inverse FFT
=
Slide: Hoiem
![Page 42: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/42.jpg)
Fourier Matlab demo
![Page 43: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/43.jpg)
FFT in Matlab• Filtering with fft
• Displaying with fft
im = double(imread(‘…'))/255; im = rgb2gray(im); % “im” should be a gray-scale floating point image[imh, imw] = size(im); hs = 50; % filter half-sizefil = fspecial('gaussian', hs*2+1, 10); fftsize = 1024; % should be order of 2 (for speed) and include paddingim_fft = fft2(im, fftsize, fftsize); % 1) fft im with paddingfil_fft = fft2(fil, fftsize, fftsize); % 2) fft fil, pad to same size as imageim_fil_fft = im_fft .* fil_fft; % 3) multiply fft imagesim_fil = ifft2(im_fil_fft); % 4) inverse fft2im_fil = im_fil(1+hs:size(im,1)+hs, 1+hs:size(im, 2)+hs); % 5) remove padding
figure(1), imagesc(log(abs(fftshift(im_fft)))), axis image, colormap jet
Slide: Hoiem
![Page 44: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/44.jpg)
Why does the Gaussian give a nice smooth image, but the square filter give edgy artifacts?
Gaussian Box filter
Filtering
![Page 45: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/45.jpg)
Gaussian
![Page 46: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/46.jpg)
Box Filter
![Page 47: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/47.jpg)
Why does a lower resolution image still make sense to us? What do we lose?
Image: http://www.flickr.com/photos/igorms/136916757/
Sampling
![Page 48: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/48.jpg)
Throw away every other row and column to create a 1/2 size image
Subsampling by a factor of 2
![Page 49: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/49.jpg)
• 1D example (sinewave):
Source: S. Marschner
Aliasing problem
![Page 50: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/50.jpg)
Source: S. Marschner
• 1D example (sinewave):
Aliasing problem
![Page 51: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/51.jpg)
• Sub-sampling may be dangerous….• Characteristic errors may appear:
– “Wagon wheels rolling the wrong way in movies”
– “Checkerboards disintegrate in ray tracing”– “Striped shirts look funny on color television”
Source: D. Forsyth
Aliasing problem
![Page 52: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/52.jpg)
Aliasing in video
Slide by Steve Seitz
![Page 53: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/53.jpg)
Source: A. Efros
Aliasing in graphics
![Page 54: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/54.jpg)
Sampling and aliasing
![Page 55: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/55.jpg)
• When sampling a signal at discrete intervals, the sampling frequency must be 2 fmax
• fmax = max frequency of the input signal
• This will allows to reconstruct the original perfectly from the sampled version
good
bad
v v v
Nyquist-Shannon Sampling Theorem
![Page 56: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/56.jpg)
Anti-aliasing
Solutions:• Sample more often
• Get rid of all frequencies that are greater than half the new sampling frequency– Will lose information– But it’s better than aliasing– Apply a smoothing filter
![Page 57: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/57.jpg)
Algorithm for downsampling by factor of 2
1. Start with image(h, w)2. Apply low-pass filter
im_blur = imfilter(image, fspecial(‘gaussian’, 7, 1))
3. Sample every other pixelim_small = im_blur(1:2:end, 1:2:end);
![Page 58: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/58.jpg)
Anti-aliasing
Forsyth and Ponce 2002
![Page 59: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/59.jpg)
Subsampling without pre-filtering
1/4 (2x zoom) 1/8 (4x zoom)1/2
Slide by Steve Seitz
![Page 60: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/60.jpg)
Subsampling with Gaussian pre-filtering
G 1/4 G 1/8Gaussian 1/2
Slide by Steve Seitz
![Page 61: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/61.jpg)
• Early processing in humans filters for various orientations and scales of frequency
• Perceptual cues in the mid-high frequencies dominate perception• When we see an image from far away, we are effectively subsampling it
Early Visual Processing: Multi-scale edge and blob filters
Clues from Human Perception
![Page 62: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/62.jpg)
Campbell-Robson contrast sensitivity curveCampbell-Robson contrast sensitivity curve
![Page 63: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/63.jpg)
Hybrid Image in FFT
Hybrid Image Low-passed Image High-passed Image
![Page 64: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/64.jpg)
Things to Remember• Sometimes it makes sense to think of
images and filtering in the frequency domain– Fourier analysis
• Can be faster to filter using FFT for large images (N logN vs. N2 for auto-correlation)
• Images are mostly smooth– Basis for compression
• Remember to low-pass before sampling
![Page 65: Thinking in Frequency Computer Vision James Hays Slides: Hoiem, Efros, and others.](https://reader036.fdocuments.us/reader036/viewer/2022070408/56649e6f5503460f94b6cc61/html5/thumbnails/65.jpg)
Questions