Single image depth-from-defocus with a learned covariance: algorithm and performance model for co-design

B. Buat; P. Trouvé-Peloux; F. Champagnat; G. Le Besnerais

doi:10.1117/12.2621252

20 May 2022 Single image depth-from-defocus with a learned covariance: algorithm and performance model for co-design

B. Buat, P. Trouvé-Peloux, F. Champagnat, G. Le Besnerais

Proceedings Volume 12136, Unconventional Optical Imaging III; 121360L (2022) https://doi.org/10.1117/12.2621252
Event: SPIE Photonics Europe, 2022, Strasbourg, France

Abstract

This paper deals with Single Image Depth-From-Defocus (SIDFD), a depth estimation approach based on local estimation of defocus blur. As both blur and scene are unknown in a single image, generic scene and blur models are commonly used in DFD algorithms. In contrast, we propose to directly learn image covariance using a limited set of calibration images which indeed encode both scene and blur (i.e. depth) information. Depth can then be estimated from a single image patch using a maximum likelihood criterion defined using the learned covariance. Here, we also propose a performance model based on the calculation of the Cram´er-Rao Bound with a learned scene model to predict the theoretical depth accuracy of SIDFD system. We validate our SIDFD algorithm and our performance model on an active chromatic SIDFD system dedicated to industrial inspection.

Conference Presentation

Citation Download Citation

B. Buat, P. Trouvé-Peloux, F. Champagnat, and G. Le Besnerais "Single image depth-from-defocus with a learned covariance: algorithm and performance model for co-design", Proc. SPIE 12136, Unconventional Optical Imaging III, 121360L (20 May 2022); https://doi.org/10.1117/12.2621252

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available