Enforcing spatially coherent structures in shape from focus

Most of the depth optimization schemes in shape from focus (SFF) enforce spatial coherence through convex energy functionals which oversmooth depth edges. Further, usually, no additional information about the scene is incorporated while estimating depth. In this work, we tackle the first issue by em...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Multimedia tools and applications Jg. 82; H. 23; S. 36431 - 36447
Hauptverfasser: Ali, Usman, Mahmood, Muhammad Tariq
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York Springer US 01.09.2023
Springer Nature B.V
Schlagworte:
Algorithms
Cameras
Coherence
Computer Communication Networks
Computer Science
Cross correlation
Data Structures and Information Theory
Design
Image sequencing
Multimedia
Multimedia Information Systems
Optimization
Regularization
Regularization methods
Shape recognition
Special Purpose and Application-Based Systems
Shape from focus
Guided image filtering
Nonconvex optimization
Focus measure
ISSN:1380-7501, 1573-7721
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Most of the depth optimization schemes in shape from focus (SFF) enforce spatial coherence through convex energy functionals which oversmooth depth edges. Further, usually, no additional information about the scene is incorporated while estimating depth. In this work, we tackle the first issue by employing a nonconvex penalty that preserves depth edges effectively. For the second issue, we design a novel guidance map for SFF based on the cross-correlation between image sequence and focus volume (FV). This cross-correlation-based guidance enforces the coherent structures between the image sequence and FV. The proposed regularization framework fuses information from the guidance map, iteratively updated depth map, and the structural similarity between them. The nonconvex objective function has been solved through the majorize-minimization algorithm. An analysis has been presented that indicates the convergence behavior of the solver. Experiments have been conducted using a variety of synthetic and real image sequences. Qualitative and quantitative comparison with state-of-the-art methods indicates that the proposed method provides better depth maps.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-023-14984-z