The orthogonal polynomial transform (OPT) can be used to effectively separate spatially correlative signals and spatially incoherent noise without losing waveform amplitude. To create a flattened dimension where the OPT can be optimally applied, we derive a plane-wave trace continuation relation for flattening the curved seismic events. Trace prediction in the flattening process and the subsequent OPT are both demonstrated to be amplitude-preserving. The robust slope estimation approach can obtain more robust performance than the state-of-the-art plane-wave destruction (PWD) method in the presence of strong noise. The proposed framework has been applied to several synthetic, field pre-stack and post-stack seismic data and are shown to better preserve the amplitude variations than other alterative methods.