Resolution-enhanced BM3D for post-stack weak signal recovery

Abstract Improving the signal-to-noise ratio (SNR) and enhancing resolution are essential for accurately recovering weak signals in seismic signal processing. Block-Matching and 3D filtering (BM3D) is a widely used denoising algorithm in data processing that employs two key stages—hard thresholding and Wiener filtering—to achieve multidimensional noise attenuation and improve SNR. However, the traditional BM3D algorithm does not fully consider the complex stratum effects present in seismic data, which limits its effectiveness in weak signal recovery. To address this limitation, we propose a resolution-enhanced BM3D (RE-BM3D), which refines both key stages to improve the recovery of weak signals. In the hard thresholding stage, we introduce an inverse filtering operator with Tikhonov regularization to better recover weak signals attenuated by stratum filtering. Additionally, a dynamic threshold is applied based on the variance of spectral coefficients to further enhance the accuracy of noise attenuation. In the Wiener filtering stage, regularized inverse filtering is employed to enhance vertical resolution, while the shrinkage coefficient is calculated based on the variance of spectral coefficients to achieve optimal SNR improvement. Experimental results show that RE-BM3D successfully enhances vertical resolution and recovers weak seismic signals, effectively overcoming the limitations of traditional BM3D in seismic data processing.