부산대학교 도서관

Bullet time is a type of visual effect commonly used in films, television and games that makes time seem to slow down or stop while still preserving dynamic details in the scene. It usually requires multiple sets of cameras to move slowly with the subject and is synthesized using post-production techniques, which is costly and one-time. The dynamic scene perspective reconstruction technology based on the neural rendering field can be used to solve this requirement. However, most current methods are poor in reconstruction accuracy due to the blurred input image and overfitting of dynamic and static regions. Based on the NSFF algorithm, we reconstruct the common time special effects scenes in movies and television from a new perspective. We add a fuzzy kernel to the network for reconstruction and analysis of the fuzzy process and input the clear perspective after analysis into NSFF to improve accuracy. We use the optical flow prediction information to suppress the dynamic network timely and force it to improve the reconstruction effect of dynamic and static scenes independently, which improves the network's ability to understand and reconstruct dynamic and static scenes. We design a new dynamic and static cross-entropy loss to solve the overfitting problem of dynamic and static scenes. Experimental results show the improved NSFF-DSC model improves the reconstruction accuracy and enhances the understanding ability of dynamic and static scenes.