Panoramic Image Quality Assessment Based on Scanpath Prediction and Bidimensional Feature Representation

Panoramic image quality assessment (PIQA) has become a high-profile visual task with the public’s high attention to immersive media. Most existing PIQA metrics neglect the consumers’ viewing behavior of perceiving panoramic image quality along a scanpath in 360° and the significance of both local and global feature representation (GFR), which hinders their further performance development. Inspired by these facts, we propose a no-reference (NR) PIQA metric based on scanpath prediction and bidimensional feature representation (SPP-BFR). The SPP module is achieved by the minimum entropy difference-based fixation point estimation method proposed by us, which has the advantages of both simulating the viewing behavior and reducing the computation overhead. The bidimensional features are explored by a large kernel attention (LKA)-based local feature representation network and the self-attention mechanism-based GFR network, respectively. Extensive experiments conducted on publicly released datasets show the superiority of the SPP-BFR metric to the state of the arts.