1 Another Look at Camera Control Karan Singh †, Cindy Grimm, Nisha Sudarsanan Media and Machines Lab Department of Computer Science and Engineering Washington.

1

Another Look at Camera Control

Karan Singh†, Cindy Grimm, Nisha Sudarsanan

Media and Machines LabDepartment of Computer Science and Engineering

Washington University in St. Louis

†University of Toronto

2Cameras

What is camera control?

Manipulation of camera parameters• Projection of 3D geometry into 2D plane

Applications• Interactively navigate around a scene (games)• Create a fixed camera path (movies)• Use different viewpoints to understand structure

(visualization)• Take pictures

3Cameras

Cameras in different fields

Computer graphics• From, at, up, field of view• View matrix

4Cameras

Cameras in different fields (cont.)

Film, photography• Camera is physical object• Describe movement of camera itself

• Dolly in/out, tilt, pan, roll• Zoom in/out

Graphics systems use same vocabulary• Through-the-lens• Exterior views

• Camera-centric• Reposition the camera

• Object-centric• Reposition the camera relative to an

object

5Cameras

Cameras in different fields (cont.)

Mathematics

• 4X4 matrix

• 11 degrees of freedom

• Straight lines to straight lines

Computer vision

• 3X4 matrix

• Don’t keep depth

Graphics maps 11 dof to “useful” parametersJ. C. Michener, I. B. Carlbom, Natural and efficient viewing parameters, SIGGRAPH 80

• Six extrinsic parameters (position, orientation)

• 5 intrinsic parameters (center-of-projection, focal length, skew, aspect ratio)

6Cameras

Cameras in different fields (cont.)

Artists have a qualitative vocabulary• Describe relationship of camera to object in the

scene• Perspective effects One point

Two point

Three point

Horizon line

Vanishing point

7Cameras

Goals

Visualization of camera parameters in the 2D image• E.g., feed back on perspective distortion

Manipulate current projection• New position indicates desired projection change• Changes appropriate camera parameters

• May change multiple parameters

Minimal mouse, keyboard use• Current approaches use entire right button• Click-through interface

8Cameras

Related work

Original camera paperJ. C. Michener, I. B. Carlbom, Natural and efficient viewing parameters, SIGGRAPH 80

Trackball manipulationMichael Chen, S. Joy Mountford, and

Abigail Sellen,A Study in Interactive 3D Rotation using 2D Input Devices , SIGGRAPH

K. Henriksen, J. Sporring, and K. Hornbaek

Virtual trackballs revisited, IEEE TVCG

9Cameras

Related work

Through-the-lens camera control

• Use image constraints to change camera parameters

• Not very stable

Jim Blinn, Where am I? What am I Looking at?, IEEE CG&A, 1988

Michael Gleicher and Andrew Witkin, Through-the-lens camera control, SIGGRAPH 92

10Cameras

The IBar

A cube centered along the look vector• Changing the rendering of the cube

changes the camera in a corresponding way

• Different segments move limbs simultaneously• Cyan = top and bottom left limbs• Red = left and right bottom limbs

• Rendering of cube reflects projection parameters

11Cameras

Demo (traditional)

Dolly in and out

Rotate left-right up-down spin

Pan (object-centric)

Zoom (object-centric)

Pan (camera-centric)

Zoom (camera-centric)

Allows framing

12Cameras

Demo (Perspective change)

Dolly + zoom

Dolly in and out

Center of projection (horizontal – vertical)

13Cameras

Demo (Just Weird)

Aspect ratio

Skew

14Cameras

Camera- versus object-centric

Camera-centric• Allows “framing” of objects• Position cube in relation to scene

Object-centric• Traditional camera-moves-with-mouse

Nice to have both• Map different limbs

• E.g., zoom using the left handle is camera-centric, the right handle is object-centric

15Cameras

Some implementation details

Camera parameters

T – Eye position

L – Look vector

V – Up vector

W,H – width, height

f – focal length

u0,v0 – center of projection

d – distance to object

V

LT

f

H

d

R

Wv0

u0

V

LTf H

16Cameras

Drawing the IBar

LssRssII

VssII

VsvRsuLdTI

fdHs

ccttl

cmt

m

00

/

Cube edge is centered on look vector•Adjust for center of projection•Size of cube adjusted so is always sc high on screen

17Cameras

Drawing the IBar (cont.)

Draw horizontal bar at horizon line

t

1-t

p1 p2

18Cameras

Drawing the IBar (cont.)

Feedback• IBar highlights when mouse is over active part• Indicate selected segment with circle

19Cameras

Manipulating the IBar

Relative, not absolute• Determine which limb, and which segment selected• Determine ratio/percentage moved• Change corresponding parameter(s) by ratio

E.g., • Multiply zoom by ratio of length change• Pan by mouse movement

VsvRsvTT

fdHs

yx

/

20Cameras

Manipulating the IBarLimb movement constrained to vertical (or horizontal)

• Shift key unconstrains

Left-right movement rotates up or downUp-down movement changes center of projection

Pan unconstrained• Shift key constrains

Stem• Shift chooses aspect ratio or skew

21Cameras

Changing parameters simultaneously

Dolly plus zoom• Calculate dolly in• Find zoom that keeps everything at a distance d

away the same size

Center of projection• Translate in reverse direction

22Cameras

Camera- versus object-centric

Camera-centric• Render cube with new camera, scene with

original camera

Object-centric• Render both cube and scene with new camera

Changing parameters• Invert operation (i.e., pan in the opposite

direction)

23Cameras

Centering the IBar on an object

Allows rotation around arbitrary point• User selects point p in scene

• Determines d• Render at p

Rotation• Translate p to origin (and camera)• Rotate• Translate back

VsvRsuLdpIm

00

24Cameras

In practice

In use in short animated film, Ryan• Used for dramatic perspective changes

25Cameras

Summary

Visualization of COP, horizon line, perspective distortion• Also at arbitrary points in the scene

One mouse button for all 11 parameters• Shift key chooses less-common action• Click through interface

• Toggle key for disabling

Usable perspective manipulation• Simultaneous editing of parameters

26Cameras

Drawbacks

Remembering which parameters go where

Visual clutter

User study comparing IBar to Maya camera• Primary conclusion: camera manipulation is hard

for both naïve and knowledgeable users • Unable to manipulate camera to match a target

scene• IBar helped people to “learn” camera manipulation

27Cameras

Future work

Reducing number of parameters on widget• Multiple widgets, quick swap between them• Similar handles

Pre-viewing of manipulation effects• What does this handle do?

Bookmarking and camera paths• Visualizing in scene

• May be out of scene