Deep Learning for Camera Autofocus

IEEE Transactions on Computational Imaging, Journal Year: 2021, Volume and Issue: 7, P. 258 - 271

Published: Jan. 1, 2021

Most digital cameras use specialized autofocus sensors, such as phase detection, lidar or ultrasound, to directly measure focus state. However, sensors increase cost and complexity without optimizing final image quality. This paper proposes a new pipeline for image-based shows that neural analysis finds 5-10x faster than traditional contrast enhancement. We achieve this by learning the direct mapping between an its position. In further with conventional methods, AI methods can generate scene-based trajectories optimize synthesized quality dynamic three dimensional scenes. propose control strategy varies focal position dynamically maximize estimated from stack. rule-based agent learned different scenarios show their advantages over other stacking methods.

Language: Английский

---

Composite Focus Measure for High Quality Depth Maps

Published: Oct. 1, 2017

Depth from focus is a highly accessible method to estimate the 3D structure of everyday scenes. Today's DSLR and mobile cameras facilitate easy capture multiple focused images scene. Focus measures (FMs) that amount at each pixel form basis depth-from-focus methods. Several FMs have been proposed in past new ones will emerge future, with their own strengths. We weighted combination standard outperforms others on wide range scene types. The resulting composite measure consists are consensus one another but not chorus. Our two-stage pipeline first estimates fine depth using measure. A cost-volume propagation step then assigns depths confident pixels others. can generate high quality maps just top five our This positive towards estimation scenes no special equipment.

Language: Английский

---

3D Imaging Based on Depth Measurement Technologies

Sensors, Journal Year: 2018, Volume and Issue: 18(11), P. 3711 - 3711

Published: Oct. 31, 2018

Three-dimensional (3D) imaging has attracted more and interest because of its widespread applications, especially in information life science. These techniques can be broadly divided into two types: ray-based wavefront-based 3D imaging. Issues such as quality system complexity these limit the applications significantly, therefore many investigations have focused on from depth measurements. This paper presents an overview measurements, provides a summary connection between techniques.

Language: Английский

---

A unilateral 3D indoor positioning system employing optical camera communications

IET Optoelectronics, Journal Year: 2023, Volume and Issue: 17(4), P. 110 - 119

Published: June 9, 2023

Abstract This article investigates the use of a visible light positioning system in an indoor environment to provide three dimensional (3D) high‐accuracy solution. The proposed leveraged single light‐emitting diode and image sensor at transmitter receiver (Rx) respectively. can retrieve 3D coordinate Rx using combination angle arrival received signal strength (RSS). To mitigate error induced by lens Rx, novel method is experimentally tested. authors show that, outperforms previously reported RSS under all circumstances it immune varying exposure times within standard range 250 µs 4 ms. demonstrate that algorithm achieve root mean squared 7.56 cm.

Language: Английский

---

A Depth From Defocus Measurement System Using a Liquid Lens Objective for Extended Depth Range

IEEE Transactions on Instrumentation and Measurement, Journal Year: 2017, Volume and Issue: 66(3), P. 441 - 450

Published: Jan. 4, 2017

A novel depth from defocus (DFD) measurement system is presented, where the extension of range performed using an emergent technology based on liquid lenses. suitable set different focal lengths, obtained by properly changing lens supply voltage, provides multiple camera settings without duplicating elements or moving parts. simple and compact setup, with a single camera/illuminator coaxial assembly, obtained. The active DFD technique modulation profilometry for estimation contrast at each image point as function range. Two methods are proposed, both combination curves, derived specific length. In first method (intensity method), information recovered directly whereas in second (differential measured curve pairs. We σ ₀ 0.55 mm over 60 intensity (0.92% total range) 0.76 135 differential (0.56% range). Thus, within state-of-the-art systems, allows, being almost equal, remarkable

Language: Английский

---

Accurate Depth Map Estimation from Small Motions

Published: Oct. 1, 2017

With the growing use of digital lightweight cameras, generating 3D information has become an important challenge in computer vision. Despite several attempts presented literature to solve this challenge, it remains open problem when comes structural accuracy depth map and required baseline (distance between first last frames) capture a sequence images. In paper, novel approach is proposed compute high quality dense together with semi-dense/dense structure from images captured on narrow baseline. Computing small motions been for decades because uncertain calculation values using - up 12mm. The method can, fact, perform much wider range baselines 8 mm 400 while respecting reference frame. evaluation done more than 10 sets recorded motion clips baseline, 7 stereo Middlebury benchmark. Preliminary results indicate that better performance terms comparison current state art methods. Also, stable even only low number frames are available processing.

Language: Английский

---

RefocusGAN: Scene Refocusing Using a Single Image

Lecture notes in computer science, Journal Year: 2018, Volume and Issue: unknown, P. 519 - 535

Published: Jan. 1, 2018

Language: Английский

---

Application of preconditioned alternating direction method of multipliers in depth from focal stack

Journal of Electronic Imaging, Journal Year: 2018, Volume and Issue: 27(02), P. 1 - 1

Published: April 6, 2018

Postcapture refocusing effect in smartphone cameras is achievable using focal stacks. However, the accuracy of this totally dependent on combination depth layers stack. The extended field application can be improved significantly by computing an accurate map, which has been open issue for decades. To tackle issue, a framework proposed based preconditioned alternating direction method multipliers from stack and synthetic defocus application. In addition to its ability provide high structural accuracy, optimization function can, fact, converge faster better than state-of-the-art methods. qualitative evaluation done 21 sets stacks compared against five other Later, 10 light image have transformed into quantitative purposes. Preliminary results indicate that performance terms comparison current

Language: Английский

---

Physically inspired depth-from-defocus

Image and Vision Computing, Journal Year: 2016, Volume and Issue: 57, P. 114 - 129

Published: Dec. 14, 2016

Language: Английский

---

Self-Supervised Single-Image Depth Estimation From Focus and Defocus Clues

IEEE Robotics and Automation Letters, Journal Year: 2021, Volume and Issue: 6(4), P. 6281 - 6288

Published: June 28, 2021

Self-supervised depth estimation has recently demonstrated promising performance compared to the supervised methods on challenging indoor scenes. However, majority of efforts mainly focus exploiting photometric and geometric consistency via forward image warping backward warping, based monocular videos or stereo pairs. The influence defocus blur is neglected, resulting in a limited for objects scenes out focus. In this work, we propose first framework simultaneous from single focal stacks using depth-from-defocus depth-from-focus algorithms. proposed network able learn optimal mapping information contained image, generate simulated stack all-in-focus train estimator an stack. addition validation our method both synthetic NYUv2 dataset real DSLR dataset, also collect own camera further verify it. Experiments demonstrate that system surpasses state-of-the-art over 4% accuracy achieves superb among without direct supervision synthesized which been rarely explored.

Language: Английский

---