TY - GEN
T1 - Task Oriented Image Quality Assessment for Synthesized Images
AU - Xu, Ke
AU - Zhang, Qian
AU - Yang, Fei
AU - Jiang, Zhanghao
AU - Lee, Boon Giin
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.
PY - 2024/12/5
Y1 - 2024/12/5
N2 - In this study, we propose a new general learning-based framework, named Task-Oriented Image Quality Assessment, for evaluating the performance of Reference-guided image synthesis (RIS) tasks. Our framework uniquely employs both source and target images to construct content- and style-encoded feature embeddings, and then evaluates the quality of the synthesized images by comparing their feature distances to those of the source and target images. We designed a two-branch network that embeds both content and style elements simultaneously. Our training process uses a style-level interpolation strategy to generate intermediate styled images for training, eliminating the need for human annotations. The quality score is calculated using a ratio-based distance that considers both the synthesized image from the source image and to the target image. Our method was evaluated using the HIDER dataset and RESIDE dataset, which provide subject scores for each image. The obtained result shows the efficiency of our method.
AB - In this study, we propose a new general learning-based framework, named Task-Oriented Image Quality Assessment, for evaluating the performance of Reference-guided image synthesis (RIS) tasks. Our framework uniquely employs both source and target images to construct content- and style-encoded feature embeddings, and then evaluates the quality of the synthesized images by comparing their feature distances to those of the source and target images. We designed a two-branch network that embeds both content and style elements simultaneously. Our training process uses a style-level interpolation strategy to generate intermediate styled images for training, eliminating the need for human annotations. The quality score is calculated using a ratio-based distance that considers both the synthesized image from the source image and to the target image. Our method was evaluated using the HIDER dataset and RESIDE dataset, which provide subject scores for each image. The obtained result shows the efficiency of our method.
KW - Image quality assessment
KW - Reference-guided image synthesis
KW - Two-branch networks
UR - http://www.scopus.com/inward/record.url?scp=85213032518&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-78389-0_10
DO - 10.1007/978-3-031-78389-0_10
M3 - Conference contribution
AN - SCOPUS:85213032518
SN - 9783031783883
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 140
EP - 153
BT - Pattern Recognition - 27th International Conference, ICPR 2024, Proceedings
A2 - Antonacopoulos, Apostolos
A2 - Chaudhuri, Subhasis
A2 - Chellappa, Rama
A2 - Liu, Cheng-Lin
A2 - Bhattacharya, Saumik
A2 - Pal, Umapada
PB - Springer Science and Business Media Deutschland GmbH
T2 - 27th International Conference on Pattern Recognition, ICPR 2024
Y2 - 1 December 2024 through 5 December 2024
ER -