The Study of Phase-Based Optical Flow Technique Using an Adaptive Bilateral Filter

Ju Hwan LEE; Sung Yun PARK; Sung Jae KIM; Sung Min KIM

doi:10.1587/transinf.E95.D.658

The Study of Phase-Based Optical Flow Technique Using an Adaptive Bilateral Filter

Ju Hwan LEE, Sung Yun PARK, Sung Jae KIM, Sung Min KIM

Full Text Views

0

Cite this

Summary :

The purpose of this study is to propose an advanced phase-based optical flow method with improved tracking accuracy for motion flow. The proposed method is mainly based on adaptive bilateral filtering (ABF) and Gabor based spatial filtering. ABF aims to preserve the maximum boundary information of the original image, while the spatial filtering aims to accurately compute the local variations. Our method tracks the optical flow in three stages. Firstly, the input images are filtered by using ABF and a spatial filter to remove noises and to preserve the maximum contour information. The component velocities are then computed based on the phase gradient of each pixel. Secondly, irregular pixels are eliminated, if the phase differences are not linear over the image frames. Lastly, the entire velocity is derived by integrating the component velocities of each pixel. In order to evaluate the tracking accuracy of the proposed method, we have examined its performance for synthetic and realistic images for which the ground truth data were known. As a result, it was observed that the proposed technique offers higher accuracy than the existing optical flow methods.

Publication: IEICE TRANSACTIONS on Information Vol.E95-D No.2 pp.658-667

Publication Date: 2012/02/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1587/transinf.E95.D.658

Type of Manuscript: PAPER

Category: Image Recognition, Computer Vision

Cite this

Copy

Ju Hwan LEE, Sung Yun PARK, Sung Jae KIM, Sung Min KIM, "The Study of Phase-Based Optical Flow Technique Using an Adaptive Bilateral Filter" in IEICE TRANSACTIONS on Information, vol. E95-D, no. 2, pp. 658-667, February 2012, doi: 10.1587/transinf.E95.D.658.
Abstract: The purpose of this study is to propose an advanced phase-based optical flow method with improved tracking accuracy for motion flow. The proposed method is mainly based on adaptive bilateral filtering (ABF) and Gabor based spatial filtering. ABF aims to preserve the maximum boundary information of the original image, while the spatial filtering aims to accurately compute the local variations. Our method tracks the optical flow in three stages. Firstly, the input images are filtered by using ABF and a spatial filter to remove noises and to preserve the maximum contour information. The component velocities are then computed based on the phase gradient of each pixel. Secondly, irregular pixels are eliminated, if the phase differences are not linear over the image frames. Lastly, the entire velocity is derived by integrating the component velocities of each pixel. In order to evaluate the tracking accuracy of the proposed method, we have examined its performance for synthetic and realistic images for which the ground truth data were known. As a result, it was observed that the proposed technique offers higher accuracy than the existing optical flow methods.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E95.D.658/_p

Copy

@ARTICLE{e95-d_2_658,
author={Ju Hwan LEE, Sung Yun PARK, Sung Jae KIM, Sung Min KIM, },
journal={IEICE TRANSACTIONS on Information},
title={The Study of Phase-Based Optical Flow Technique Using an Adaptive Bilateral Filter},
year={2012},
volume={E95-D},
number={2},
pages={658-667},
abstract={The purpose of this study is to propose an advanced phase-based optical flow method with improved tracking accuracy for motion flow. The proposed method is mainly based on adaptive bilateral filtering (ABF) and Gabor based spatial filtering. ABF aims to preserve the maximum boundary information of the original image, while the spatial filtering aims to accurately compute the local variations. Our method tracks the optical flow in three stages. Firstly, the input images are filtered by using ABF and a spatial filter to remove noises and to preserve the maximum contour information. The component velocities are then computed based on the phase gradient of each pixel. Secondly, irregular pixels are eliminated, if the phase differences are not linear over the image frames. Lastly, the entire velocity is derived by integrating the component velocities of each pixel. In order to evaluate the tracking accuracy of the proposed method, we have examined its performance for synthetic and realistic images for which the ground truth data were known. As a result, it was observed that the proposed technique offers higher accuracy than the existing optical flow methods.},
keywords={},
doi={10.1587/transinf.E95.D.658},
ISSN={1745-1361},
month={February},}

Copy

TY - JOUR
TI - The Study of Phase-Based Optical Flow Technique Using an Adaptive Bilateral Filter
T2 - IEICE TRANSACTIONS on Information
SP - 658
EP - 667
AU - Ju Hwan LEE
AU - Sung Yun PARK
AU - Sung Jae KIM
AU - Sung Min KIM
PY - 2012
DO - 10.1587/transinf.E95.D.658
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E95-D
IS - 2
JA - IEICE TRANSACTIONS on Information
Y1 - February 2012
AB - The purpose of this study is to propose an advanced phase-based optical flow method with improved tracking accuracy for motion flow. The proposed method is mainly based on adaptive bilateral filtering (ABF) and Gabor based spatial filtering. ABF aims to preserve the maximum boundary information of the original image, while the spatial filtering aims to accurately compute the local variations. Our method tracks the optical flow in three stages. Firstly, the input images are filtered by using ABF and a spatial filter to remove noises and to preserve the maximum contour information. The component velocities are then computed based on the phase gradient of each pixel. Secondly, irregular pixels are eliminated, if the phase differences are not linear over the image frames. Lastly, the entire velocity is derived by integrating the component velocities of each pixel. In order to evaluate the tracking accuracy of the proposed method, we have examined its performance for synthetic and realistic images for which the ground truth data were known. As a result, it was observed that the proposed technique offers higher accuracy than the existing optical flow methods.
ER -