Video-based Reconstruction of Animatable Human Characters

by raja on January 23, 2011

in Assignment 1

Video-based Reconstruction of Animatable Human Characters (assigned)

Sentence
We present a new performance capture approach that incorporates a physically-based cloth model to reconstruct a rigged fully-animatable virtual double of a real person in loose apparel from multi-view video recordings. (1st line of abstract)

Problem
Performances captured via approaches based on multi-view recordings that enable reconstruction of detailed motion and time-varying geometry of humans are limited to playbacks; their motions and deformations cannot be edited by an animator.

Key Idea
Construction of the “reference performance” that comprises of a deforming mesh sequence that represents the detailed surface geometry of the moving actor as well as a sequence of joint angle parameters, which is then decomposed into rigid body parts and non-rigid pieces of the apparel.

What the paper lets you do
Reconstruct a fully-animatable model of a person in general apparel using multi-view video (cameras around a circle)

What it could be used for
Faster character animation creation time (even real time!)
Capture real life events (using the required camera arrangement) and turn them into game levels quickly, how cool is that!

Other materials
Project web page & video

Previous post:

Next post: