28
Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams mages/Demo from Steve Seitz, Wikipedia

Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Embed Size (px)

Citation preview

Page 1: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Photometric Image Formation

CSE 559: Computer VisionGuest Lecturer: Austin Abrams

Images/Demo from Steve Seitz, Wikipedia

Page 2: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

How are images made?

• One half: geometric vision– “how the pixel projected onto the image”

• Today: photometric vision (aka radiometric)– “how the pixel got its color”

Page 3: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Vision and Graphics

Properties of a sceneImage

Computer Graphics

Vision

Page 4: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Image Formation Approach

• Come up with a model for how the scene was created

• Given images, find the most likely properties that fit that model

Page 5: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Diffuse Surfaces

Brightness of a pixel depends on:• object color• lighting direction• surface normal

But NOT view direction!

Page 6: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Lambertian Cosine Law

• The intensity of an observed diffuse object is proportional to the cosine of the angle between the normal and lighting direction

= ρ L N

I = ρ cos θ

= ρ |L||N| cos θL Nθ

Page 7: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

=

L N = L N

Page 8: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

= x

I = ρ L N

Page 9: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Recovering Albedo and Normals

• Can you decompose a single image into its albedo and normal images?

Page 10: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

=

x

x

x

Page 11: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Photometric Stereo

• Given multiple images taken with varying illumination, recover albedo and normals.– take pictures in dark room with varying

illumination.– estimate lighting directions L.– recover albedo and normals.

Page 12: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Side note 1: How to get the lighting direction?

• Put a shiny sphere in the scene• Sphere’s geometry (normals) are known• Find specular highlight

Page 13: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Side-note 2: Why “Stereo”?

Surface normals provide constraints on depth differences

Page 14: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Photometric Stereo

• If L is known, and albedo is grayscale this is a linear problem.

I = ρ(L N) = ρ (Lx Nx + Ly Ny + Lz Nz ) = Lx Nxρ + Ly Nyρ + Lz Nzρ = Lx a + Ly b + Lz c

Page 15: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Lx1 Ly1 Lz1

Lx2 Ly2 Lz2

Lx3 Ly3 Lz3

Lxn Lyn Lzn

I1I2I3…In

abc

=

I = ρ(L N) = Lx a + Ly b + Lz c

Then:ρ = sqrt(a2 + b2 + c2)N = (a,b,c) / ρ

For each pixel:

Page 16: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Demo

Page 17: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

When does this model fail?

I ≠ ρ (L N)

Page 18: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Attached shadows

I = ρ max(L N, 0)

L N > 0

L N = 0

L N < 0

Page 19: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Cast Shadows, Ambient Light

I = ρ (S L N + a) S = 0 or 1

Page 20: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Radiometric Camera Calibration

• Pixel intensities are usually not proportional to the energy that hit the CCD

RAW image Published image

Page 21: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Radiometric Camera Calibration

f

RAW

Published

Page 22: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Radiometric Camera Calibration

Observed = f(RAW)

(Grossberg and Nayar)

f -1 (Observed) = RAW

Page 23: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Radiometric Camera Calibration

• How do you model f -1?

f -1(x) = xγ

f -1(x) = c0 + c1x + c2x2 + c3x3 + …

f -1(x) = f0(x) + f1(x) c1 + f2(x)c2 + …

mean camera curve basis camera curves

Page 24: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Radiometric Camera Calibration

I = f (ρ (S L N + a))

Adding exposure:

I = f (e ρ (S L N + a))

Page 25: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Heliometric StereoGiven lots of images from a stable webcam,

use lighting from the sun to recover:

I = f (e ρ (S L N + a))

Page 26: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Heliometric Stereo

Page 27: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Heliometric Stereo

Page 28: Photometric Image Formation CSE 559: Computer Vision Guest Lecturer: Austin Abrams Images/Demo from Steve Seitz, Wikipedia

Heliometric Stereo