Quantitative quality assessment of stitched panoramic images

摘要

In this study, the authors present quantitative quality assessment indices for measuring geometric and photometric qualities of stitched panoramic images, since both geometric and photometric qualities are important for obtaining a seamless panoramic image. For geometric quality assessment, our approach is based on the use of structural similarity (SSIM) index between the high-frequency information (HFI) of both geometrically corrected unstitched images and the stitched panoramic image. This measure is called 'HFI_SSIM'. The complementary low-frequency information from the same pair of images is used for assessing the photometric quality of the stitched image using the spectral angle mapper and intensity magnitude ratio measures. The proposed quality metrics are tested on both synthetic and true datasets comprising of indoor and outdoor scenes. The results obtained, highlight the interest of using the proposed quantitative measures for assessing the quality of stitched panoramic images.