Syllabus for 15869,
ImageBased Modeling and Rendering
Fall 1999
Professors: Steve Seitz and Paul Heckbert
Computer Science Department
Carnegie Mellon University
In topical order (which is only approximately chronological).

Introduction

what are imagebased techniques?

administrative matters

video examples

Images (2D)

image warping

transformations: rigid body, affine, projective, other

brief review of homogeneous coordinates

resampling, bilinear interpolation

algorithms for image warping: forward, backward, twopass

image mosaics

applications: remote sensing of earth and planets, panoramas,
superresolution, video

projective mosaics: planar scenes and panoramas

environment maps: cylinder, sphere, cube

Quicktime VR

image registration for mosaicing

manual methods

automatic methods:
sum of squared differences, mutual information

image compositing

alpha blending

feathering for image mosaics

morphing

correspondence specification

BeierNeely method

gridbased techniques

interactive techniques

finding image contours with "snakes"

matte extraction,
"scissors"

environment matting

Depth Images (2.5D)

zbuffer, range data, disparity

view interpolation

review of perspective camera model

ChenWilliams algorithm

occlusion, hole filling problem

plane + parallax

Macmillan's backtofront rendering algorithm, plenoptic modeling

sprites (images in layers)

automatic extraction of image layers from video

layered depth images (LDI's)

the (proposed) Talisman rendering architecture

impostors (resampling geometric models)

3D Modeling

cameras

optics of real cameras

projective geometry (pointline duality)

view morphing (Seitz)

representing geometry:
meshes,
voxels,
point clouds

texture mapping

applications: reflectance, bumps, transparency, specularity, environment

antialiasing and texture filtering

acquiring geometry

camera calibration

photogrammetry

stereo

correspondence problem,
epipolar geometry

stereo algorithms,
occlusion,
hole filling problem

structure from motion maybe

voxel and silhouette models

volume intersection

finding silhouettes (again)

voxel coloring (Seitz & Dyer)

active methods maybe

merging range scans maybe

representing radiance

the manydimensional nature of radiance

typical limitation to diffuse, opaque surfaces

bidirectional reflectance distribution functions (BRDF's)

environment maps

computing radiance/BRDF's

ray tracing

high dynamic range images

compositing real and synthetic images

Raybased representations (4D)

light fields: acquiring, compression, rendering (Levoy, Gortler)

Shum's concentric mosaics

Applications

Faces

modelbased techniques: Blanz & Vetter

learning techniques

special effects:
Dayton Taylor's "timetrack"

teleconferencing maybe

image and video rewrite

Motion maybe
15869
Steve Seitz and Paul Heckbert, Sept. 1999