We propose a feature design method for a mobile visual search based on binary features and a bag-of-visual words framework. In mobile visual search, detection error and quantization error are unavoidable due to viewpoint changes and cause performance degradation. Typical approaches to visual search extract features from a single view of reference images, though such features are insufficient to manage detection and quantization errors. In this paper, we extract features from multiview synthetic images. These features are selected according to our novel reliability measure which enables robust recognition against various viewpoint changes. We regard feature selection as a maximum coverage problem. That is, we find a finite set of features maximizing an objective function under certain constraints. As this problem is NP-hard and thus computationally infeasible, we explore approximate solutions based on a greedy algorithm. For this purpose, we propose novel constraint functions which are designed to be consistent with the match conditions in the visual search method. Experiments show that the proposed method improves retrieval accuracy by 12.7 percentage points without increasing the database size or changing the search procedure. In other words, the proposed method enables more accurate search without adversely affecting the database size, computational cost, and memory requirement.
Kohei MATSUZAKI
KDDI Research, Inc.
Kazuyuki TASAKA
KDDI Research, Inc.
Hiromasa YANAGIHARA
KDDI Research, Inc.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Kohei MATSUZAKI, Kazuyuki TASAKA, Hiromasa YANAGIHARA, "Local Feature Reliability Measure Consistent with Match Conditions for Mobile Visual Search" in IEICE TRANSACTIONS on Information,
vol. E101-D, no. 12, pp. 3170-3180, December 2018, doi: 10.1587/transinf.2018EDP7107.
Abstract: We propose a feature design method for a mobile visual search based on binary features and a bag-of-visual words framework. In mobile visual search, detection error and quantization error are unavoidable due to viewpoint changes and cause performance degradation. Typical approaches to visual search extract features from a single view of reference images, though such features are insufficient to manage detection and quantization errors. In this paper, we extract features from multiview synthetic images. These features are selected according to our novel reliability measure which enables robust recognition against various viewpoint changes. We regard feature selection as a maximum coverage problem. That is, we find a finite set of features maximizing an objective function under certain constraints. As this problem is NP-hard and thus computationally infeasible, we explore approximate solutions based on a greedy algorithm. For this purpose, we propose novel constraint functions which are designed to be consistent with the match conditions in the visual search method. Experiments show that the proposed method improves retrieval accuracy by 12.7 percentage points without increasing the database size or changing the search procedure. In other words, the proposed method enables more accurate search without adversely affecting the database size, computational cost, and memory requirement.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2018EDP7107/_p
Copy
@ARTICLE{e101-d_12_3170,
author={Kohei MATSUZAKI, Kazuyuki TASAKA, Hiromasa YANAGIHARA, },
journal={IEICE TRANSACTIONS on Information},
title={Local Feature Reliability Measure Consistent with Match Conditions for Mobile Visual Search},
year={2018},
volume={E101-D},
number={12},
pages={3170-3180},
abstract={We propose a feature design method for a mobile visual search based on binary features and a bag-of-visual words framework. In mobile visual search, detection error and quantization error are unavoidable due to viewpoint changes and cause performance degradation. Typical approaches to visual search extract features from a single view of reference images, though such features are insufficient to manage detection and quantization errors. In this paper, we extract features from multiview synthetic images. These features are selected according to our novel reliability measure which enables robust recognition against various viewpoint changes. We regard feature selection as a maximum coverage problem. That is, we find a finite set of features maximizing an objective function under certain constraints. As this problem is NP-hard and thus computationally infeasible, we explore approximate solutions based on a greedy algorithm. For this purpose, we propose novel constraint functions which are designed to be consistent with the match conditions in the visual search method. Experiments show that the proposed method improves retrieval accuracy by 12.7 percentage points without increasing the database size or changing the search procedure. In other words, the proposed method enables more accurate search without adversely affecting the database size, computational cost, and memory requirement.},
keywords={},
doi={10.1587/transinf.2018EDP7107},
ISSN={1745-1361},
month={December},}
Copy
TY - JOUR
TI - Local Feature Reliability Measure Consistent with Match Conditions for Mobile Visual Search
T2 - IEICE TRANSACTIONS on Information
SP - 3170
EP - 3180
AU - Kohei MATSUZAKI
AU - Kazuyuki TASAKA
AU - Hiromasa YANAGIHARA
PY - 2018
DO - 10.1587/transinf.2018EDP7107
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E101-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2018
AB - We propose a feature design method for a mobile visual search based on binary features and a bag-of-visual words framework. In mobile visual search, detection error and quantization error are unavoidable due to viewpoint changes and cause performance degradation. Typical approaches to visual search extract features from a single view of reference images, though such features are insufficient to manage detection and quantization errors. In this paper, we extract features from multiview synthetic images. These features are selected according to our novel reliability measure which enables robust recognition against various viewpoint changes. We regard feature selection as a maximum coverage problem. That is, we find a finite set of features maximizing an objective function under certain constraints. As this problem is NP-hard and thus computationally infeasible, we explore approximate solutions based on a greedy algorithm. For this purpose, we propose novel constraint functions which are designed to be consistent with the match conditions in the visual search method. Experiments show that the proposed method improves retrieval accuracy by 12.7 percentage points without increasing the database size or changing the search procedure. In other words, the proposed method enables more accurate search without adversely affecting the database size, computational cost, and memory requirement.
ER -