Image-Based View Synthesis

We present a new method for rendering novel images of flexible 3D objects from a small number of example images in correspondence. The strength of the method is the ability to synthesize images whose viewing position is significantly far away from the viewing cone of the example images ("...

Full description

Bibliographic Details
Main Authors: Avidan, Shai, Evgeniou, Theodoros, Shashua, Amnon, Poggio, Tomaso
Language:en_US
Published: 2004
Subjects:
Online Access:http://hdl.handle.net/1721.1/7179
_version_ 1826204107825741824
author Avidan, Shai
Evgeniou, Theodoros
Shashua, Amnon
Poggio, Tomaso
author_facet Avidan, Shai
Evgeniou, Theodoros
Shashua, Amnon
Poggio, Tomaso
author_sort Avidan, Shai
collection MIT
description We present a new method for rendering novel images of flexible 3D objects from a small number of example images in correspondence. The strength of the method is the ability to synthesize images whose viewing position is significantly far away from the viewing cone of the example images ("view extrapolation"), yet without ever modeling the 3D structure of the scene. The method relies on synthesizing a chain of "trilinear tensors" that governs the warping function from the example images to the novel image, together with a multi-dimensional interpolation function that synthesizes the non-rigid motions of the viewed object from the virtual camera position. We show that two closely spaced example images alone are sufficient in practice to synthesize a significant viewing cone, thus demonstrating the ability of representing an object by a relatively small number of model images --- for the purpose of cheap and fast viewers that can run on standard hardware.
first_indexed 2024-09-23T12:49:01Z
id mit-1721.1/7179
institution Massachusetts Institute of Technology
language en_US
last_indexed 2024-09-23T12:49:01Z
publishDate 2004
record_format dspace
spelling mit-1721.1/71792019-04-10T11:52:40Z Image-Based View Synthesis Avidan, Shai Evgeniou, Theodoros Shashua, Amnon Poggio, Tomaso Image Based Rendering Trilinear Tensor Multidimensional Morphing We present a new method for rendering novel images of flexible 3D objects from a small number of example images in correspondence. The strength of the method is the ability to synthesize images whose viewing position is significantly far away from the viewing cone of the example images ("view extrapolation"), yet without ever modeling the 3D structure of the scene. The method relies on synthesizing a chain of "trilinear tensors" that governs the warping function from the example images to the novel image, together with a multi-dimensional interpolation function that synthesizes the non-rigid motions of the viewed object from the virtual camera position. We show that two closely spaced example images alone are sufficient in practice to synthesize a significant viewing cone, thus demonstrating the ability of representing an object by a relatively small number of model images --- for the purpose of cheap and fast viewers that can run on standard hardware. 2004-10-20T20:48:52Z 2004-10-20T20:48:52Z 1997-01-01 AIM-1603 CBCL-145 http://hdl.handle.net/1721.1/7179 en_US AIM-1603 CBCL-145 22 p. 11491930 bytes 2474047 bytes application/postscript application/pdf application/postscript application/pdf
spellingShingle Image Based Rendering
Trilinear Tensor
Multidimensional Morphing
Avidan, Shai
Evgeniou, Theodoros
Shashua, Amnon
Poggio, Tomaso
Image-Based View Synthesis
title Image-Based View Synthesis
title_full Image-Based View Synthesis
title_fullStr Image-Based View Synthesis
title_full_unstemmed Image-Based View Synthesis
title_short Image-Based View Synthesis
title_sort image based view synthesis
topic Image Based Rendering
Trilinear Tensor
Multidimensional Morphing
url http://hdl.handle.net/1721.1/7179
work_keys_str_mv AT avidanshai imagebasedviewsynthesis
AT evgenioutheodoros imagebasedviewsynthesis
AT shashuaamnon imagebasedviewsynthesis
AT poggiotomaso imagebasedviewsynthesis