Perceptual Blind Panoramic Image Quality Assessment Based on Super-Pixel

Blind objective quality assessment of panoramic images (PIQA) is a great challenge to perform highly consistent with human perception without the original panoramic images. In this paper, we propose a perceptual blind PIQA method based super-pixel, which exploits the equirectangular projection (ERP) and human perception characteristics for panoramic image to boost up the quality assessment performance. In particular, in order to make use of the local features of panoramic image, panoramic weights based on super-pixel is designed by combining ERP format and human perception. In addition, we propose panoramic-weighted structural features to predict the visual quality of panoramic images, which can reflect spherical quality accurately. Finally, we fuse and map extracted features into quality scores by applying support vector regression (SVR). The experiments demonstrate the effectiveness and superiority of our proposed metric compared with state-of-the-art PIQA methods on the public panoramic image datasets.