Results of the summarizing process for the arXiv paper: 2303.08050v1

Computer graphics images (CGIs) are widely used in various scenarios such as games and streaming media. However, the quality of CGIs often suffers from poor rendering during production and compression artifacts during transmission. Despite this challenge, few works have been dedicated to addressing computer graphics image quality assessment (CGIQA). Most image quality assessment (IQA) metrics are developed for natural scene images (NSIs) and validated on databases consisting of NSIs with synthetic distortions, which are not suitable for in-the-wild CGIs. To bridge the gap between evaluating the quality of NSIs and CGIs, this paper proposes a large-scale in-the-wild CGIQA database consisting of 6,000 CGIs called CGIQA-6k. The database covers a wider range of resolutions from 480p to 4K and contains more diverse content as well as authentic distortions. A well-organized subjective experiment is carried out in a controlled laboratory environment to obtain accurate perceptual ratings of the CGIs. Furthermore, an effective deep learning-based no-reference IQA model is proposed by utilizing multi-stage feature fusion strategy and multi-stage channel attention mechanism. The proposed model takes advantage of inter-channel information from low-level to high-level since CGIs have apparent patterns as well as rich interactive semantic content. The experimental results show that the proposed method outperforms all other state-of-the-art NR IQA methods on the constructed CGIQA-6k database and other CGIQA related databases. The major contributions of this work include constructing the largest in-the wild CGIQA database with subjective quality scores covering a wider range of resolutions and diverse content along with proposing an effective deep learning based NR IQA model through multi stage feature fusion and multi stage channel attention mechanisms. The database along with code will be released to facilitate further research in this area.

• Computer graphics images (CGIs) are widely used in various scenarios such as games and streaming media.
• The quality of CGIs often suffers from poor rendering during production and compression artifacts during transmission.
• Few works have been dedicated to addressing computer graphics image quality assessment (CGIQA).
• Most image quality assessment (IQA) metrics are developed for natural scene images (NSIs) and validated on databases consisting of NSIs with synthetic distortions, which are not suitable for in-the-wild CGIs.
• To bridge the gap between evaluating the quality of NSIs and CGIs, this paper proposes a large-scale in-the-wild CGIQA database consisting of 6,000 CGIs called CGIQA-6k.
• The proposed model takes advantage of inter-channel information from low-level to high-level since CGIs have apparent patterns as well as rich interactive semantic content.
• The experimental results show that the proposed method outperforms all other state-of-the-art NR IQA methods on the constructed CGIQA-6k database and other CGIQA related databases.
• The major contributions of this work include constructing the largest in-the wild CGIQA database with subjective quality scores covering a wider range of resolutions and diverse content along with proposing an effective deep learning based NR IQA model through multi stage feature fusion and multi stage channel attention mechanisms.
• The database along with code will be released to facilitate further research in this area.

Computer Graphics Image Quality Assessment: A Comprehensive Overview

Computer graphics images (CGIs) are widely used in various scenarios such as games and streaming media. However, the quality of CGIs often suffers from poor rendering during production and compression artifacts during transmission. Despite this challenge, few works have been dedicated to addressing computer graphics image quality assessment (CGIQA). In this article, we will provide a comprehensive overview of CGIQA including its challenges, existing databases and models for CGIQA evaluation.

Challenges with CGIQA

The main challenge with CGIQA is that most image quality assessment (IQA) metrics are developed for natural scene images (NSIs) and validated on databases consisting of NSIs with synthetic distortions, which are not suitable for in-the-wild CGIs. To bridge the gap between evaluating the quality of NSIs and CGIs, it is necessary to construct an in-the-wild database that covers a wider range of resolutions from 480p to 4K containing more diverse content along with authentic distortions.

Existing Databases & Models

To address these challenges, this paper proposes a large-scale in-the-wild CGIQA database consisting of 6,000 CGIs called CGIQA-6k. A well-organized subjective experiment is carried out in a controlled laboratory environment to obtain accurate perceptual ratings of the CGIs. Furthermore, an effective deep learning based no reference IQA model is proposed by utilizing multi stage feature fusion strategy and multi stage channel attention mechanism. The proposed model takes advantage of inter channel information from low level to high level since CGIs have apparent patterns as well as rich interactive semantic content. The experimental results show that the proposed method outperforms all other state-of-the art NR IQA methods on the constructed CGIQA 6k database and other related databases.

Conclusion

In conclusion, this work has made major contributions towards constructing an in wild CGIQA database covering a wider range of resolutions from 480p to 4K containing more diverse content along with authentic distortions as well as proposing an effective deep learning based NR IQA model through multi stage feature fusion and multi stage channel attention mechanisms. The database along with code will be released to facilitate further research in this area making it easier for researchers to evaluate their models on real world data sets instead of relying solely on synthetic data sets which may not accurately reflect real world scenarios or performance metrics .