Computer graphics images (CGIs) are artificially generated by means of computer programs and are widely perceived under various scenarios, such as games, streaming media, etc. In practical, the quality of CGIs consistently suffers from poor rendering during the production and inevitable compression artifacts during the transmission of multimedia applications. However, few works have been dedicated to dealing with the challenge of computer graphics images quality assessment (CGIQA). Most image quality assessment (IQA) metrics are developed for natural scene images (NSIs) and validated on the 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, we construct a large-scale in-the-wild CGIQA database consisting of 6,000 CGIs (CGIQA-6k) and carry out the subjective experiment in a well-controlled laboratory environment to obtain the accurate perceptual ratings of the CGIs. Then, we propose an effective deep learning-based no-reference (NR) IQA model by utilizing multi-stage feature fusion strategy and multi-stage channel attention mechanism. The major motivation of the proposed model is to make full use of inter-channel information from low-level to high-level since CGIs have apparent patterns as well as rich interactive semantic content. 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 database along with the code will be released to facilitate further research.

翻译：计算机图形图像（CGIs）是通过计算机程序人工生成的图像，在游戏、流媒体等多种场景中被广泛感知。在实际应用中，CGIs的质量在多媒体应用的生产过程中常因渲染不佳而受损，并在传输过程中遭受不可避免的压缩伪影影响。然而，目前鲜有研究专门应对计算机图形图像质量评估（CGIQA）这一挑战。大多数图像质量评估（IQA）指标是为自然场景图像（NSIs）开发的，并在包含合成失真的NSIs数据库上验证，这些方法不适用于野外CGIs。为弥合NSIs与CGIs质量评估之间的差距，我们构建了一个包含6,000张CGIs的大规模野外CGIQA数据库（CGIQA-6k），并在严格控制的实验室环境中进行主观实验，以获取CGIs的精准感知评分。随后，我们提出了一种基于深度学习的有效无参考（NR）IQA模型，采用多阶段特征融合策略和多阶段通道注意力机制。该模型的主要动机在于充分利用从低层到高层的通道间信息，因为CGIs具有明显模式及丰富的交互式语义内容。实验结果表明，在构建的CGIQA-6k数据库及其他CGIQA相关数据库上，所提方法优于所有其他最先进的NR IQA方法。该数据库及代码将公开，以促进进一步研究。