Half of an image is enough for quality assessment

Junyong You; Yuan Lin; Jari Korhonen

doi:10.48550/arXiv.2301.12891

Half of an image is enough for quality assessment

Junyong You, Yuan Lin, Jari Korhonen

Research output: Chapter in Book/Report/Conference proceeding › Published conference contribution

Abstract

Deep networks have demonstrated promising results in the field of Image Quality Assessment (IQA). However, there has been limited research on understanding how deep models in IQA work. This study introduces a novel positional masked transformer for IQA and provides insights into the contribution of different regions of an image towards its overall quality. Results indicate that half of an image may play a trivial role in determining image quality, while the other half is critical. This observation is extended to several other CNN-based IQA models, revealing that half of the image regions can significantly impact the overall image quality. To further enhance our understanding, three semantic measures (saliency, frequency, and objectness) were derived and found to have high correlation with the importance of image regions in IQA.

Original language	English
Title of host publication	2023 IEEE International Conference on Image Processing (ICIP)
Publisher	IEEE Press
Pages	61-65
Number of pages	5
ISBN (Electronic)	978-1-7281-9835-4
DOIs	https://doi.org/10.48550/arXiv.2301.12891 https://doi.org/10.1109/ICIP49359.2023.10223124
Publication status	Published - Oct 2023
Event	2023 IEEE International Conference on Image Processing - Kuala Lumpur, Malaysia Duration: 8 Oct 2023 → 11 Oct 2023 https://2023.ieeeicip.org/

Conference

Conference	2023 IEEE International Conference on Image Processing
Abbreviated title	ICIP 2023
Country/Territory	Malaysia
City	Kuala Lumpur
Period	8/10/23 → 11/10/23
Internet address	https://2023.ieeeicip.org/

Keywords

Explainable AI (XAI)
image quality assessment (IQA)
positional masking
semantic measures

Access to Document

10.48550/arXiv.2301.12891Licence: CC BY
10.1109/ICIP49359.2023.10223124Licence: Unspecified

Cite this

@inproceedings{a36d94d7ab4442abaf52ee51386c3c31,

title = "Half of an image is enough for quality assessment",

abstract = "Deep networks have demonstrated promising results in the field of Image Quality Assessment (IQA). However, there has been limited research on understanding how deep models in IQA work. This study introduces a novel positional masked transformer for IQA and provides insights into the contribution of different regions of an image towards its overall quality. Results indicate that half of an image may play a trivial role in determining image quality, while the other half is critical. This observation is extended to several other CNN-based IQA models, revealing that half of the image regions can significantly impact the overall image quality. To further enhance our understanding, three semantic measures (saliency, frequency, and objectness) were derived and found to have high correlation with the importance of image regions in IQA.",

keywords = "Explainable AI (XAI), image quality assessment (IQA), positional masking, semantic measures",

author = "Junyong You and Yuan Lin and Jari Korhonen",

year = "2023",

month = oct,

doi = "10.48550/arXiv.2301.12891",

language = "English",

pages = "61--65",

booktitle = "2023 IEEE International Conference on Image Processing (ICIP)",

publisher = "IEEE Press",

note = "2023 IEEE International Conference on Image Processing, ICIP 2023 ; Conference date: 08-10-2023 Through 11-10-2023",

url = "https://2023.ieeeicip.org/",

}

TY - GEN

T1 - Half of an image is enough for quality assessment

AU - You, Junyong

AU - Lin, Yuan

AU - Korhonen, Jari

PY - 2023/10

Y1 - 2023/10

N2 - Deep networks have demonstrated promising results in the field of Image Quality Assessment (IQA). However, there has been limited research on understanding how deep models in IQA work. This study introduces a novel positional masked transformer for IQA and provides insights into the contribution of different regions of an image towards its overall quality. Results indicate that half of an image may play a trivial role in determining image quality, while the other half is critical. This observation is extended to several other CNN-based IQA models, revealing that half of the image regions can significantly impact the overall image quality. To further enhance our understanding, three semantic measures (saliency, frequency, and objectness) were derived and found to have high correlation with the importance of image regions in IQA.

AB - Deep networks have demonstrated promising results in the field of Image Quality Assessment (IQA). However, there has been limited research on understanding how deep models in IQA work. This study introduces a novel positional masked transformer for IQA and provides insights into the contribution of different regions of an image towards its overall quality. Results indicate that half of an image may play a trivial role in determining image quality, while the other half is critical. This observation is extended to several other CNN-based IQA models, revealing that half of the image regions can significantly impact the overall image quality. To further enhance our understanding, three semantic measures (saliency, frequency, and objectness) were derived and found to have high correlation with the importance of image regions in IQA.

KW - Explainable AI (XAI)

KW - image quality assessment (IQA)

KW - positional masking

KW - semantic measures

U2 - 10.48550/arXiv.2301.12891

DO - 10.48550/arXiv.2301.12891

M3 - Published conference contribution

SP - 61

EP - 65

BT - 2023 IEEE International Conference on Image Processing (ICIP)

PB - IEEE Press

T2 - 2023 IEEE International Conference on Image Processing

Y2 - 8 October 2023 through 11 October 2023

ER -

Half of an image is enough for quality assessment

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this