<p>Seismic signal denoising and feature extraction play crucial roles in geophysics, particularly in applications like earthquake monitoring, subsurface imaging, and reservoir characterization. However, the inherent noise in seismic data poses significant challenges for accurately identifying seismic arrival times. Traditional methods such as STA/LTA and P-phase picker have proven effective in simple scenarios but struggle in complex and noisy environments. This study adapts and extends the Convex Non-Convex Fused Lasso Signal Approximation (CNC-FLSA) framework, originally proposed for biomedical signal processing, to seismic data analysis. By optimizing key parameters and leveraging non-convex regularization, the CNC-FLSA method achieves superior denoising and precise seismic event detection. Comprehensive evaluations on synthetic and real datasets demonstrate its robustness in diverse noise conditions and its advantages over traditional methods and state-of-the-art techniques. The proposed method not only improves accuracy but also offers computational efficiency, making it a viable tool for seismic signal processing applications.</p>