Accuracy improvement of depth estimation with tilted optics by optimizing neural network

Hiroshi Ikeoka, Takayuki Hamamoto

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

1 Citation (Scopus)

Abstract

We have been investigating a novel depth estimation system that adopts tilted-lens optics for real-time usage, e.g., automotive tasks. Herein, we obtained depth values for each pixel from the sharpness ratio of only two tilted optics images; we used a monocular camera system with a spectroscopic mirror. However, the method causes some estimation errors because of the difference between the optical theory and the actual camera system. Therefore, to reduce the error, we adopted a neural network to obtain the depth map. In this paper, we report our improvement by optimizing the neural network construction which calculates the depth value for each pixel from 3 × 3 pixel values at each image and y-coordinate.

Original languageEnglish
Title of host publicationInternational Workshop on Advanced Image Technology, IWAIT 2019
EditorsKazuya Hayase, Qian Kemao, Phooi Yee Lau, Wen-Nung Lie, Lu Yu, Sanun Srisuk, Yung-Lyul Lee
PublisherSPIE
ISBN (Electronic)9781510627734
DOIs
Publication statusPublished - 1 Jan 2019
EventInternational Workshop on Advanced Image Technology 2019, IWAIT 2019 - Singapore, Singapore
Duration: 6 Jan 20199 Jan 2019

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume11049
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceInternational Workshop on Advanced Image Technology 2019, IWAIT 2019
CountrySingapore
CitySingapore
Period6/01/199/01/19

Fingerprint

Depth Estimation
Optics
Pixel
Pixels
optics
Neural Networks
Neural networks
pixels
Camera
Cameras
Depth Map
cameras
Sharpness
Estimation Error
Error analysis
Lens
Lenses
Mirror
sharpness
Mirrors

Keywords

  • blur
  • deep learning
  • defocus
  • depth estimation
  • distance estimation
  • neural network
  • tilted optics

Cite this

Ikeoka, H., & Hamamoto, T. (2019). Accuracy improvement of depth estimation with tilted optics by optimizing neural network. In K. Hayase, Q. Kemao, P. Y. Lau, W-N. Lie, L. Yu, S. Srisuk, & Y-L. Lee (Eds.), International Workshop on Advanced Image Technology, IWAIT 2019 [1104934] (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11049). SPIE. https://doi.org/10.1117/12.2521101
Ikeoka, Hiroshi ; Hamamoto, Takayuki. / Accuracy improvement of depth estimation with tilted optics by optimizing neural network. International Workshop on Advanced Image Technology, IWAIT 2019. editor / Kazuya Hayase ; Qian Kemao ; Phooi Yee Lau ; Wen-Nung Lie ; Lu Yu ; Sanun Srisuk ; Yung-Lyul Lee. SPIE, 2019. (Proceedings of SPIE - The International Society for Optical Engineering).
@inproceedings{bb153ae7c05c48758f11349acf850dbb,
title = "Accuracy improvement of depth estimation with tilted optics by optimizing neural network",
abstract = "We have been investigating a novel depth estimation system that adopts tilted-lens optics for real-time usage, e.g., automotive tasks. Herein, we obtained depth values for each pixel from the sharpness ratio of only two tilted optics images; we used a monocular camera system with a spectroscopic mirror. However, the method causes some estimation errors because of the difference between the optical theory and the actual camera system. Therefore, to reduce the error, we adopted a neural network to obtain the depth map. In this paper, we report our improvement by optimizing the neural network construction which calculates the depth value for each pixel from 3 × 3 pixel values at each image and y-coordinate.",
keywords = "blur, deep learning, defocus, depth estimation, distance estimation, neural network, tilted optics",
author = "Hiroshi Ikeoka and Takayuki Hamamoto",
year = "2019",
month = "1",
day = "1",
doi = "10.1117/12.2521101",
language = "English",
series = "Proceedings of SPIE - The International Society for Optical Engineering",
publisher = "SPIE",
editor = "Kazuya Hayase and Qian Kemao and Lau, {Phooi Yee} and Wen-Nung Lie and Lu Yu and Sanun Srisuk and Yung-Lyul Lee",
booktitle = "International Workshop on Advanced Image Technology, IWAIT 2019",

}

Ikeoka, H & Hamamoto, T 2019, Accuracy improvement of depth estimation with tilted optics by optimizing neural network. in K Hayase, Q Kemao, PY Lau, W-N Lie, L Yu, S Srisuk & Y-L Lee (eds), International Workshop on Advanced Image Technology, IWAIT 2019., 1104934, Proceedings of SPIE - The International Society for Optical Engineering, vol. 11049, SPIE, International Workshop on Advanced Image Technology 2019, IWAIT 2019, Singapore, Singapore, 6/01/19. https://doi.org/10.1117/12.2521101

Accuracy improvement of depth estimation with tilted optics by optimizing neural network. / Ikeoka, Hiroshi; Hamamoto, Takayuki.

International Workshop on Advanced Image Technology, IWAIT 2019. ed. / Kazuya Hayase; Qian Kemao; Phooi Yee Lau; Wen-Nung Lie; Lu Yu; Sanun Srisuk; Yung-Lyul Lee. SPIE, 2019. 1104934 (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11049).

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

TY - GEN

T1 - Accuracy improvement of depth estimation with tilted optics by optimizing neural network

AU - Ikeoka, Hiroshi

AU - Hamamoto, Takayuki

PY - 2019/1/1

Y1 - 2019/1/1

N2 - We have been investigating a novel depth estimation system that adopts tilted-lens optics for real-time usage, e.g., automotive tasks. Herein, we obtained depth values for each pixel from the sharpness ratio of only two tilted optics images; we used a monocular camera system with a spectroscopic mirror. However, the method causes some estimation errors because of the difference between the optical theory and the actual camera system. Therefore, to reduce the error, we adopted a neural network to obtain the depth map. In this paper, we report our improvement by optimizing the neural network construction which calculates the depth value for each pixel from 3 × 3 pixel values at each image and y-coordinate.

AB - We have been investigating a novel depth estimation system that adopts tilted-lens optics for real-time usage, e.g., automotive tasks. Herein, we obtained depth values for each pixel from the sharpness ratio of only two tilted optics images; we used a monocular camera system with a spectroscopic mirror. However, the method causes some estimation errors because of the difference between the optical theory and the actual camera system. Therefore, to reduce the error, we adopted a neural network to obtain the depth map. In this paper, we report our improvement by optimizing the neural network construction which calculates the depth value for each pixel from 3 × 3 pixel values at each image and y-coordinate.

KW - blur

KW - deep learning

KW - defocus

KW - depth estimation

KW - distance estimation

KW - neural network

KW - tilted optics

UR - http://www.scopus.com/inward/record.url?scp=85063888846&partnerID=8YFLogxK

U2 - 10.1117/12.2521101

DO - 10.1117/12.2521101

M3 - Conference contribution

AN - SCOPUS:85063888846

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - International Workshop on Advanced Image Technology, IWAIT 2019

A2 - Hayase, Kazuya

A2 - Kemao, Qian

A2 - Lau, Phooi Yee

A2 - Lie, Wen-Nung

A2 - Yu, Lu

A2 - Srisuk, Sanun

A2 - Lee, Yung-Lyul

PB - SPIE

ER -

Ikeoka H, Hamamoto T. Accuracy improvement of depth estimation with tilted optics by optimizing neural network. In Hayase K, Kemao Q, Lau PY, Lie W-N, Yu L, Srisuk S, Lee Y-L, editors, International Workshop on Advanced Image Technology, IWAIT 2019. SPIE. 2019. 1104934. (Proceedings of SPIE - The International Society for Optical Engineering). https://doi.org/10.1117/12.2521101