Compressive acquisition of dynamic scenes

Aswin C. Sankaranarayanan, Pavan Turaga, Richard G. Baraniuk, Rama Chellappa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

52 Citations (Scopus)

Abstract

Compressive sensing (CS) is a new approach for the acquisition and recovery of sparse signals and images that enables sampling rates significantly below the classical Nyquist rate. Despite significant progress in the theory and methods of CS, little headway has been made in compressive video acquisition and recovery. Video CS is complicated by the ephemeral nature of dynamic events, which makes direct extensions of standard CS imaging architectures and signal models infeasible. In this paper, we develop a new framework for video CS for dynamic textured scenes that models the evolution of the scene as a linear dynamical system (LDS). This reduces the video recovery problem to first estimating the model parameters of the LDS from compressive measurements, from which the image frames are then reconstructed. We exploit the low-dimensional dynamic parameters (the state sequence) and high-dimensional static parameters (the observation matrix) of the LDS to devise a novel compressive measurement strategy that measures only the dynamic part of the scene at each instant and accumulates measurements over time to estimate the static parameters. This enables us to considerably lower the compressive measurement rate considerably. We validate our approach with a range of experiments including classification experiments that highlight the effectiveness of the proposed approach.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages129-142
Number of pages14
Volume6311 LNCS
EditionPART 1
DOIs
StatePublished - 2010
Externally publishedYes
Event11th European Conference on Computer Vision, ECCV 2010 - Heraklion, Crete, Greece
Duration: Sep 5 2010Sep 11 2010

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume6311 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other11th European Conference on Computer Vision, ECCV 2010
CountryGreece
CityHeraklion, Crete
Period9/5/109/11/10

Fingerprint

Compressive Sensing
Linear Dynamical Systems
Dynamical systems
Recovery
Accumulate
Instant
Experiments
Experiment
Sampling
Imaging techniques
High-dimensional
Imaging
Acquisition
Model
Estimate
Range of data

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Sankaranarayanan, A. C., Turaga, P., Baraniuk, R. G., & Chellappa, R. (2010). Compressive acquisition of dynamic scenes. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 1 ed., Vol. 6311 LNCS, pp. 129-142). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6311 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-15549-9_10

Compressive acquisition of dynamic scenes. / Sankaranarayanan, Aswin C.; Turaga, Pavan; Baraniuk, Richard G.; Chellappa, Rama.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6311 LNCS PART 1. ed. 2010. p. 129-142 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6311 LNCS, No. PART 1).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Sankaranarayanan, AC, Turaga, P, Baraniuk, RG & Chellappa, R 2010, Compressive acquisition of dynamic scenes. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 edn, vol. 6311 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 6311 LNCS, pp. 129-142, 11th European Conference on Computer Vision, ECCV 2010, Heraklion, Crete, Greece, 9/5/10. https://doi.org/10.1007/978-3-642-15549-9_10
Sankaranarayanan AC, Turaga P, Baraniuk RG, Chellappa R. Compressive acquisition of dynamic scenes. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 ed. Vol. 6311 LNCS. 2010. p. 129-142. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-642-15549-9_10
Sankaranarayanan, Aswin C. ; Turaga, Pavan ; Baraniuk, Richard G. ; Chellappa, Rama. / Compressive acquisition of dynamic scenes. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6311 LNCS PART 1. ed. 2010. pp. 129-142 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).
@inproceedings{36c4e6be57494142ab38e5e5decf0ae0,
title = "Compressive acquisition of dynamic scenes",
abstract = "Compressive sensing (CS) is a new approach for the acquisition and recovery of sparse signals and images that enables sampling rates significantly below the classical Nyquist rate. Despite significant progress in the theory and methods of CS, little headway has been made in compressive video acquisition and recovery. Video CS is complicated by the ephemeral nature of dynamic events, which makes direct extensions of standard CS imaging architectures and signal models infeasible. In this paper, we develop a new framework for video CS for dynamic textured scenes that models the evolution of the scene as a linear dynamical system (LDS). This reduces the video recovery problem to first estimating the model parameters of the LDS from compressive measurements, from which the image frames are then reconstructed. We exploit the low-dimensional dynamic parameters (the state sequence) and high-dimensional static parameters (the observation matrix) of the LDS to devise a novel compressive measurement strategy that measures only the dynamic part of the scene at each instant and accumulates measurements over time to estimate the static parameters. This enables us to considerably lower the compressive measurement rate considerably. We validate our approach with a range of experiments including classification experiments that highlight the effectiveness of the proposed approach.",
author = "Sankaranarayanan, {Aswin C.} and Pavan Turaga and Baraniuk, {Richard G.} and Rama Chellappa",
year = "2010",
doi = "10.1007/978-3-642-15549-9_10",
language = "English (US)",
isbn = "3642155480",
volume = "6311 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 1",
pages = "129--142",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 1",

}

TY - GEN

T1 - Compressive acquisition of dynamic scenes

AU - Sankaranarayanan, Aswin C.

AU - Turaga, Pavan

AU - Baraniuk, Richard G.

AU - Chellappa, Rama

PY - 2010

Y1 - 2010

N2 - Compressive sensing (CS) is a new approach for the acquisition and recovery of sparse signals and images that enables sampling rates significantly below the classical Nyquist rate. Despite significant progress in the theory and methods of CS, little headway has been made in compressive video acquisition and recovery. Video CS is complicated by the ephemeral nature of dynamic events, which makes direct extensions of standard CS imaging architectures and signal models infeasible. In this paper, we develop a new framework for video CS for dynamic textured scenes that models the evolution of the scene as a linear dynamical system (LDS). This reduces the video recovery problem to first estimating the model parameters of the LDS from compressive measurements, from which the image frames are then reconstructed. We exploit the low-dimensional dynamic parameters (the state sequence) and high-dimensional static parameters (the observation matrix) of the LDS to devise a novel compressive measurement strategy that measures only the dynamic part of the scene at each instant and accumulates measurements over time to estimate the static parameters. This enables us to considerably lower the compressive measurement rate considerably. We validate our approach with a range of experiments including classification experiments that highlight the effectiveness of the proposed approach.

AB - Compressive sensing (CS) is a new approach for the acquisition and recovery of sparse signals and images that enables sampling rates significantly below the classical Nyquist rate. Despite significant progress in the theory and methods of CS, little headway has been made in compressive video acquisition and recovery. Video CS is complicated by the ephemeral nature of dynamic events, which makes direct extensions of standard CS imaging architectures and signal models infeasible. In this paper, we develop a new framework for video CS for dynamic textured scenes that models the evolution of the scene as a linear dynamical system (LDS). This reduces the video recovery problem to first estimating the model parameters of the LDS from compressive measurements, from which the image frames are then reconstructed. We exploit the low-dimensional dynamic parameters (the state sequence) and high-dimensional static parameters (the observation matrix) of the LDS to devise a novel compressive measurement strategy that measures only the dynamic part of the scene at each instant and accumulates measurements over time to estimate the static parameters. This enables us to considerably lower the compressive measurement rate considerably. We validate our approach with a range of experiments including classification experiments that highlight the effectiveness of the proposed approach.

UR - http://www.scopus.com/inward/record.url?scp=78149333918&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78149333918&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-15549-9_10

DO - 10.1007/978-3-642-15549-9_10

M3 - Conference contribution

SN - 3642155480

SN - 9783642155482

VL - 6311 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 129

EP - 142

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -