Aiding self-supervised coherent noise suppression by the introduction of signal segmentation using blind-spot networks

Blind-spot networks have been shown to be natural noise suppressors under the assumption that noise is unpredictable based on the information fed into the network during training. Trained in a self-supervised manner, such approaches only utilise the original raw data to determine to remove the noise. In this work, we propose two novel elements for enhancing blind-spot denoising: 1) the introduction of a 2-class segmentation task to aid the network in identification of interest areas of signals that require particular attention during denoising, and; 2) the introduction of a trace-wise noise mask designed to obscure the coherency of noise from being observed by the network. The joint scheme is achieved by introducing a joint loss function to balance between the two deep learning tasks. As such, the final joint scheme is the combination of a self-supervised, blind-spot denoising procedure and a supervised segmentation procedure. We illustrate how the joint scheme can improve the denoising performance of the network, hypothesising that this is due to the introduction of prior information guiding the denoising procedure to areas of focus. Preliminary results from synthetic data contaminated by trace-wise noise, show an increase in the structural similarity index from 0.989 to 0.995, when comparing the optimal joint-scheme versus the pure denoising procedure. Future work will extend the procedure to field data where rule-based approaches will be used to generate the segmentation labels.