steve seitz dept. computer science & eng. university of washington 3d photography: beyond...

36
Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Post on 20-Jan-2016

231 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Steve SeitzDept. Computer Science & Eng.

University of Washington

3D Photography:Beyond Perspective

Page 2: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Perspective Projection

Humans evolved with perspective eyes• Capture light along rays that converge at a single point

Page 3: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Cameras also evolved with perspective optics

• Optimized for humans, not computers!

Page 4: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

A non-perspective image

Page 5: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective
Page 6: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

The Blue Marble, NASA satellite image composite

Page 7: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

The Blue Marble, NASA satellite image composite

Page 8: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

The Blue Marble, NASA satellite image composite

Page 9: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Cyberware Scanner

Page 10: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Print Gallery, by M.C. Escher, 1956

Page 11: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Panorama from Disney’s 1940 film Pinocchio (from Wood et al., SIGGRAPH 1997)

Page 12: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective
Page 13: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

What’s an Image?

An image is any 2D subset of rays in space• Actually, the light energy flowing along these rays• Any 2D “slice” of the plenoptic function

perspective image general image

Page 14: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Non-perspective Imaging

Issues:• What other types of images are possible?• Which images are useful?• How can we capture these images?

Page 15: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Path Images

Page 16: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Path Images

Page 17: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Path Images

x

y

t

Page 18: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Path Images

x

y

t

Page 19: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Path Images

Pushbroom images• satellite

Bolles et al. [87]• EPI

Tsuji et al. [92]• omni-directional image

Peleg et al. [97]• manifold mosaic

Radamacher & Bishop [99]• MCOP

. . .

x

y

t

Page 20: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Linear Path

input pushbroom images

Video Cube Demo• application by Michael Cohen et al., Microsoft

Page 21: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Circular Path

panorama(“concentric mosaic”)

circular EPI

input image

Page 22: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Circular Path

input cyclographsvideo cube

Page 23: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

What are these images good for?• Applications to computer graphics,

computer vision?

Page 24: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Circular Stereo• [Ishiguro, Yamamoto, Tsuji, 92]

• [Peleg and Ben-Ezra, 99]

• [Shum, Kalai, Seitz, 99]

• [Nayar and Karmarkar, 00]

I

y

x x0 +n

x0 -n x

0

Page 25: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Stereo Panorama

Disparity map result

Page 26: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Stereo Panorama

Dark--close, light--far

Page 27: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Circular Stereo

Advantages• 360 degree scene reconstruction• Uniform accuracy, optimal

Page 28: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Pushbroom Stereo

y

x tx0 +n

x0 -n x

0

Page 29: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Stereo Cyclographs

y

x x0 +n

x0 -n x

0

Page 30: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective
Page 31: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective
Page 32: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Stereo Cyclograph Reconstruction

Computed from two cyclograph images• Using unmodified stereo matcher [Zitnick & Kanade]

Page 33: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Stereo Path Images

Do these all produce stereo pairs?• two images with horizontal parallax

Yes

Page 34: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Stereo Path Images

How about this path? No• Yes if the camera path is a conic

– line, circle, ellipse, parabola, hyperbola

• Must capture rays lying on doubly-ruled quadrics– [Padja 2001], [Seitz 2001]

Page 35: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Stereo Parabolic Panoramas

perspective image

parabolic panorama

Page 36: Steve Seitz Dept. Computer Science & Eng. University of Washington 3D Photography: Beyond Perspective

Beyond PerspectiveCameras for humans, not machines!

• Need to rethink cameras• Image should suit the task

Future: cameras will evolve like CPU’s• First: task-specific cameras• Then: programmable cameras

– FPGA programmable camera arrays

Thanks to Jiwon Kim, Michael Cohen