No-Reference Image Quality Assessment for distorted images has always been a challenging problem due to im- age content variance and distortion diversity. Previous IQA models mostly encode explicit single-quality features of synthetic images to obtain quality-aware representations for quality score prediction. However, performance de- creases when facing real-world distortion and restored im- ages from restoration models. The reason is that they do not consider the degradation factors of the low-quality im- ages adequately. To address this issue, we first introduce the DRI method to obtain degradation vectors and qual- ity vectors of images, which separately model the degra- dation and quality information of low-quality images. After that, we add the restoration network to provide the MOS score predictor with degradation information. Then, we design the Representation-based Semantic Loss (RS Loss) to assist in enhancing effective interaction between repre- sentations. Extensive experimental results demonstrate that the proposed method performs favorably against existing state-of-the-art models on both synthetic and real-world datasets.
View on arXiv