Directional adaptive mode total variation for seismic data denoising

Seismic noise attenuation is a critical task in geophysical data processing. However, addressing directional features while preserving the curvilinear nature of complex seismic events remains a significant challenge. A model capable of handling dominant multidirectional data would be highly applicable. While directional total variation (DTV) effectively handles data with a single dominant direction, it falls short when multiple directions are present. To address this limitation, we propose a novel method for denoising seismic data that accommodates multiple dominant directions by adapting an architecture named directional adaptive mode total variation (DAM-TV). We formulate the model via convex optimization, incorporating spatially varying directional modes in total variation. This approach decomposes the two-dimensional seismic data into k distinct modes, which are then processed using an advanced mathematical algorithm to handle multiple dominant directions effectively. Additionally, a canny edge detector is used to help distinguish between the broken curve edges of noisy and denoised traces. The directional TV per mode with gradient optimization, higher SNR because of mode-wise directional smoothing, edge preservation due to adaptive DTV with Canny guided evaluation, have been executed, which made the proposed algorithm distinct from others. The experiments were conducted on both synthetic and real seismic data, ensuring that the proposed approach is more promising than existing methods.