GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS

Kede Ma; Zhengfang Duanmu; Zhou Wang

doi:10.1109/ICASSP.2018.8462176

GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS

Kede Ma
, Zhengfang Duanmu
, Zhou Wang

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

20 Citations (Scopus)

Abstract

Most existing full-reference (FR) image quality assessment (IQA) models assume that the reference and distorted images are perfectly aligned, and fail dramatically when the assumption does not hold. In this study, we first show that pre-registration, especially feature-based (as opposed to area-based) registration, is effective at reducing the performance drop of FR-IQA models. However, registration is an expensive process that often slows down the speed of the IQA algorithms by several orders of magnitude. This motivates us to construct an end-to-end convolutional neural network (CNN) for direct image quality prediction, which contains built-in invariance to geometric distortions. Our results show that when the training images are augmented by their geometrically transformed versions, the learned network performs at a high level without image registration, resulting in a fast and effective approach for geometric transformation invariant IQA.

Original language	English
Title of host publication	2018 IEEE International Conference on Acoustics, Speech, and Signal Processing
Subtitle of host publication	PROCEEDINGS
Publisher	IEEE
Pages	6732-6736
ISBN (Electronic)	9781538646588
ISBN (Print)	9781538646595
DOIs	https://doi.org/10.1109/ICASSP.2018.8462176
Publication status	Published - Apr 2018
Externally published	Yes
Event	2018 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2018) - Calgary Telus Convention Center, Calgary, Canada Duration: 15 Apr 2018 → 20 Apr 2018 https://2018.ieeeicassp.org/

Publication series

Name	International Conference on Acoustics, Speech, and Signal Processing (ICASSP)
Publisher	IEEE
ISSN (Print)	1520-6149
ISSN (Electronic)	2379-190X

Conference

Conference	2018 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2018)
Abbreviated title	ICASSP 2018
Place	Canada
City	Calgary
Period	15/04/18 → 20/04/18
Internet address	https://2018.ieeeicassp.org/

Research Keywords

Convolutional neural networks
Data augmentation
Geometric transformations
Image quality assessment
Image registration

Access Output

10.1109/ICASSP.2018.8462176

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Ma, K., Duanmu, Z., & Wang, Z. (2018). GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS. In 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing: PROCEEDINGS (pp. 6732-6736). Article 8462176 (International Conference on Acoustics, Speech, and Signal Processing (ICASSP)). IEEE. https://doi.org/10.1109/ICASSP.2018.8462176

@inproceedings{e05502e7d81f4258934dac7304c30e2f,

title = "GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS",

abstract = "Most existing full-reference (FR) image quality assessment (IQA) models assume that the reference and distorted images are perfectly aligned, and fail dramatically when the assumption does not hold. In this study, we first show that pre-registration, especially feature-based (as opposed to area-based) registration, is effective at reducing the performance drop of FR-IQA models. However, registration is an expensive process that often slows down the speed of the IQA algorithms by several orders of magnitude. This motivates us to construct an end-to-end convolutional neural network (CNN) for direct image quality prediction, which contains built-in invariance to geometric distortions. Our results show that when the training images are augmented by their geometrically transformed versions, the learned network performs at a high level without image registration, resulting in a fast and effective approach for geometric transformation invariant IQA.",

keywords = "Convolutional neural networks, Data augmentation, Geometric transformations, Image quality assessment, Image registration",

author = "Kede Ma and Zhengfang Duanmu and Zhou Wang",

year = "2018",

month = apr,

doi = "10.1109/ICASSP.2018.8462176",

language = "English",

isbn = "9781538646595",

series = "International Conference on Acoustics, Speech, and Signal Processing (ICASSP)",

publisher = "IEEE",

pages = "6732--6736",

booktitle = "2018 IEEE International Conference on Acoustics, Speech, and Signal Processing",

address = "United States",

note = "2018 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2018), ICASSP 2018 ; Conference date: 15-04-2018 Through 20-04-2018",

url = "https://2018.ieeeicassp.org/",

}

Ma, K, Duanmu, Z & Wang, Z 2018, GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS. in 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing: PROCEEDINGS., 8462176, International Conference on Acoustics, Speech, and Signal Processing (ICASSP), IEEE, pp. 6732-6736, 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2018), Calgary, Alberta, Canada, 15/04/18. https://doi.org/10.1109/ICASSP.2018.8462176

GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS. / Ma, Kede; Duanmu, Zhengfang; Wang, Zhou.
2018 IEEE International Conference on Acoustics, Speech, and Signal Processing: PROCEEDINGS. IEEE, 2018. p. 6732-6736 8462176 (International Conference on Acoustics, Speech, and Signal Processing (ICASSP)).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS

AU - Ma, Kede

AU - Duanmu, Zhengfang

AU - Wang, Zhou

PY - 2018/4

Y1 - 2018/4

N2 - Most existing full-reference (FR) image quality assessment (IQA) models assume that the reference and distorted images are perfectly aligned, and fail dramatically when the assumption does not hold. In this study, we first show that pre-registration, especially feature-based (as opposed to area-based) registration, is effective at reducing the performance drop of FR-IQA models. However, registration is an expensive process that often slows down the speed of the IQA algorithms by several orders of magnitude. This motivates us to construct an end-to-end convolutional neural network (CNN) for direct image quality prediction, which contains built-in invariance to geometric distortions. Our results show that when the training images are augmented by their geometrically transformed versions, the learned network performs at a high level without image registration, resulting in a fast and effective approach for geometric transformation invariant IQA.

AB - Most existing full-reference (FR) image quality assessment (IQA) models assume that the reference and distorted images are perfectly aligned, and fail dramatically when the assumption does not hold. In this study, we first show that pre-registration, especially feature-based (as opposed to area-based) registration, is effective at reducing the performance drop of FR-IQA models. However, registration is an expensive process that often slows down the speed of the IQA algorithms by several orders of magnitude. This motivates us to construct an end-to-end convolutional neural network (CNN) for direct image quality prediction, which contains built-in invariance to geometric distortions. Our results show that when the training images are augmented by their geometrically transformed versions, the learned network performs at a high level without image registration, resulting in a fast and effective approach for geometric transformation invariant IQA.

KW - Convolutional neural networks

KW - Data augmentation

KW - Geometric transformations

KW - Image quality assessment

KW - Image registration

UR - https://www.scopus.com/pages/publications/85054221601

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85054221601&origin=recordpage

U2 - 10.1109/ICASSP.2018.8462176

DO - 10.1109/ICASSP.2018.8462176

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 9781538646595

T3 - International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

SP - 6732

EP - 6736

BT - 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing

PB - IEEE

T2 - 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2018)

Y2 - 15 April 2018 through 20 April 2018

ER -

GEOMETRIC TRANSFORMATION INVARIANT IMAGE QUALITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS

Abstract

Publication series

Conference

Research Keywords

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this