Image-based rendering (IBR) technique generates novel views by utilizing input images captured from various viewpoints to create an immersive video experience. However, current learning-based IBR methods have limitations as they only work at the still image level, and they do not maintain consistency between consecutive frames, leading to temporal noise. To address this, we propose an intra-only framework that identifies parts of input images causing temporal artifacts in synthesized views. Our method produces better and more stable novel views for immersive video transmission. We conclude that
our framework is capable of detecting and correcting spatial features in still image level that produce artifacts in the temporal dimension.