Fast Bayesian blind deconvolution with Huber Super Gaussian priors
We report expectation Maximization (EM) based inference has already proven to be a very powerful tool to solve blind image deconvolution (BID) problems. Unfortunately, three important problems still impede the application of EM in BID: the undesirable saddle points and local minima caused by highly...
Saved in:
Published in | Digital signal processing Vol. 60; no. C |
---|---|
Main Authors | , , , , |
Format | Journal Article |
Language | English |
Published |
United States
Elsevier
06.09.2016
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Abstract | We report expectation Maximization (EM) based inference has already proven to be a very powerful tool to solve blind image deconvolution (BID) problems. Unfortunately, three important problems still impede the application of EM in BID: the undesirable saddle points and local minima caused by highly nonconvex priors, the instability around zero of some of the most interesting sparsity promoting priors, and the intrinsic high computational cost of the corresponding BID algorithm. In this paper we first show how Super Gaussian priors can be made numerically tractable around zero by introducing the family of Huber Super Gaussian priors and then present a fast EM based blind deconvolution method formulated in the image space. In the proposed computational approach, image and kernel estimation are performed by using the Alternating Direction Method of Multipliers (ADMM), which allows to exploit the advantages of FFT computation. For highly nonconvex priors, we propose a Smooth ADMM (SADMM) approach to avoid poor BID estimates. In conclusion, extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art BID methods in terms of quality of the reconstructions and speed. |
---|---|
AbstractList | We report expectation Maximization (EM) based inference has already proven to be a very powerful tool to solve blind image deconvolution (BID) problems. Unfortunately, three important problems still impede the application of EM in BID: the undesirable saddle points and local minima caused by highly nonconvex priors, the instability around zero of some of the most interesting sparsity promoting priors, and the intrinsic high computational cost of the corresponding BID algorithm. In this paper we first show how Super Gaussian priors can be made numerically tractable around zero by introducing the family of Huber Super Gaussian priors and then present a fast EM based blind deconvolution method formulated in the image space. In the proposed computational approach, image and kernel estimation are performed by using the Alternating Direction Method of Multipliers (ADMM), which allows to exploit the advantages of FFT computation. For highly nonconvex priors, we propose a Smooth ADMM (SADMM) approach to avoid poor BID estimates. In conclusion, extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art BID methods in terms of quality of the reconstructions and speed. |
Author | Zhou, Xu Vega, Miguel Molina, Rafael Zhou, Fugen Katsaggelos, Aggelos K. |
Author_xml | – sequence: 1 orcidid: 0000000161525941 fullname: Zhou, Xu organization: Beihang University, Beijing (China); ShanghaiTech University, Shanghai (China)] (ORCID:0000000161525941 – sequence: 2 fullname: Vega, Miguel organization: University de Granada, Granada (Spain) – sequence: 3 fullname: Zhou, Fugen organization: Beihang University, Beijing (China) – sequence: 4 fullname: Molina, Rafael organization: University de Granada, Granada (Spain) – sequence: 5 orcidid: 0000000345540070 fullname: Katsaggelos, Aggelos K. organization: Northwestern University, Evanston, IL (United States)] (ORCID:0000000345540070 |
BackLink | https://www.osti.gov/servlets/purl/1487466$$D View this record in Osti.gov |
BookMark | eNqNjEEOgjAUBRuDiaDeoXFPUmgB3WJE9ronpdTwDWkJv9V4e9F4ADdvZjF5EQmMNXpBwoQdslhwzoOPZ0mcMiZWJEK8M8YKkeYhKSuJjpbypRGkoe0ApqOdVtY87OAdWEOf4Hpa-1ZP9OLHec_S47ceJ7ATbsjyJgfU2x_XZFedrsc6tuigQQVOq34-NFq5JhH7QuQ5_yt6A578PPQ |
ContentType | Journal Article |
CorporateAuthor | Northwestern Univ., Evanston, IL (United States) |
CorporateAuthor_xml | – name: Northwestern Univ., Evanston, IL (United States) |
DBID | OIOZB OTOTI |
DatabaseName | OSTI.GOV - Hybrid OSTI.GOV |
DatabaseTitleList | |
DeliveryMethod | fulltext_linktorsrc |
Discipline | Engineering |
EISSN | 1095-4333 |
ExternalDocumentID | 1487466 |
GroupedDBID | --K --M .DC .~1 0R~ 1B1 1~. 1~5 4.4 457 4G. 5GY 5VS 7-5 71M 8P~ 9JN AACTN AAEDT AAEDW AAIAV AAIKJ AAKOC AALMO AALRI AAOAW AAQFI AAXUO ABBOA ABMAC ABPIF ABPTK ABYKQ ACDAQ ACGFS ACRLP ACZNC ADBBV ADEZE ADTZH AEBSH AECPX AEKER AENEX AFKWA AFTJW AGHFR AGUBO AGYEJ AHHHB AHJVU AHZHX AIALX AIEXJ AIKHN AITUG AJBFU AJOXV ALMA_UNASSIGNED_HOLDINGS AMFUW AMRAJ AOUOD AXJTR BJAXD BKOJK BLXMC CS3 DM4 DU5 EBS EFBJH EFJIC EJD EO8 EO9 EP2 EP3 F5P FDB FIRID FNPLU FYGXN G-Q GBLVA GBOLZ IHE J1W JJJVA KOM LG5 LG9 LY7 M41 MO0 N9A O-L O9- OAUVE OIOZB OTOTI OZT P-8 P-9 P2P PC. Q38 RIG ROL RPZ SDF SDG SDP SES SPC SPCBC SST SSV SSZ T5K ZMT ZU3 ~G- |
ID | FETCH-osti_scitechconnect_14874663 |
ISSN | 1051-2004 |
IngestDate | Fri May 19 00:39:28 EDT 2023 |
IsOpenAccess | true |
IsPeerReviewed | true |
IsScholarly | true |
Issue | C |
Language | English |
LinkModel | OpenURL |
MergedId | FETCHMERGED-osti_scitechconnect_14874663 |
Notes | NA0002520 USDOE National Nuclear Security Administration (NNSA) |
ORCID | 0000000161525941 0000000345540070 |
OpenAccessLink | https://www.osti.gov/servlets/purl/1487466 |
ParticipantIDs | osti_scitechconnect_1487466 |
PublicationCentury | 2000 |
PublicationDate | 2016-09-06 |
PublicationDateYYYYMMDD | 2016-09-06 |
PublicationDate_xml | – month: 09 year: 2016 text: 2016-09-06 day: 06 |
PublicationDecade | 2010 |
PublicationPlace | United States |
PublicationPlace_xml | – name: United States |
PublicationTitle | Digital signal processing |
PublicationYear | 2016 |
Publisher | Elsevier |
Publisher_xml | – name: Elsevier |
SSID | ssj0007426 |
Score | 4.2797256 |
Snippet | We report expectation Maximization (EM) based inference has already proven to be a very powerful tool to solve blind image deconvolution (BID) problems.... |
SourceID | osti |
SourceType | Open Access Repository |
SubjectTerms | blind deconvolution image deblurring image restoration MATHEMATICS AND COMPUTING variational Bayesian |
Title | Fast Bayesian blind deconvolution with Huber Super Gaussian priors |
URI | https://www.osti.gov/servlets/purl/1487466 |
Volume | 60 |
hasFullText | 1 |
inHoldings | 1 |
isFullTextHit | |
isPrint | |
link | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnZ1LT8JAEMc3ykkPxmdU1GyMN1JCKS30qCgSEzyh4Ua6jxISQwm0Jnrwszszu6XVGKJeNmRp2sKv7DyY_w5jVwFEHQLMthOFbuS0Yuk7YIYbjvKlK2IJTwS16Rw8Bv2n1sPIHxUtL0ldkoq6fP9RV_IfqjAHXFEl-weyq5PCBLwGvjACYRh_xbgXLSG8j940KSEFeIyqpjDCfbWXNWnWfib0ApaIOYz3Ubako-eLaWL_ybG-6e10gh1EaljSgfosIyHITRsll5MMkYyyfOZZm7zsYDrJ9Mv343rZpBCaDag7EBGN4rxQ36Yb3IDqqYIiB2YNdmnJhJ-1oUIWxc6FPmqxvPI6a_oG2OepW9ifVVUgRGbtVhBssk3PxerM-kdRrQMRPAnF8ouBBU1gDSz5AsNdtmOdeH5tiOyxDT3bZ9ulrR0P2A2y4TkbTmz4FzYc2XBiw4kNz9lww-aQXfbuht2-g7cwBlcF99uVWJgk07H9FN4Rq8ySmT5mPPZ83ZRt3cHtxVTYEaGSSkjVkM3Y6zT1CauuOdHp2nerbKugdMYq6SLT5-AmpeKCvsBP7j4fyQ |
link.rule.ids | 230,315,786,790,891 |
linkProvider | Elsevier |
openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Fast+Bayesian+blind+deconvolution+with+Huber+Super+Gaussian+priors&rft.jtitle=Digital+signal+processing&rft.au=Zhou%2C+Xu&rft.au=Vega%2C+Miguel&rft.au=Zhou%2C+Fugen&rft.au=Molina%2C+Rafael&rft.date=2016-09-06&rft.pub=Elsevier&rft.issn=1051-2004&rft.eissn=1095-4333&rft.volume=60&rft.issue=C&rft.externalDocID=1487466 |
thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1051-2004&client=summon |
thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1051-2004&client=summon |
thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1051-2004&client=summon |