Photoacoustic signal refinement with nonlinear energy modeling and statistical inference

Conventional linear filters often blur microvascular details in photoacoustic imaging (PAI), where broadband noise and clutter degrade signal fidelity. We introduce a training-free framework combining a modified Teager–Kaiser Energy Operator (mTKEO, Ψ_mod) with a Hidden Markov Model (HMM). The mTKEO accentuates localized energy bursts from vascular features, while the HMM adaptively segments signal states using statistical metrics to guide contrast enhancement. A final beamforming and normalization step preserves both structure and dynamic range. Evaluated on 2D k-Wave phantoms (10–30 dB SNR) and in vivo mouse-ear data, our method yields up to 8.3 dB PSNR gain, 0.0646 SSIM improvement, and 0.31 dB CNR boost versus Wiener and Kalman filters. This first mTKEO–HMM integration offers a lightweight, robust alternative to deep-learning–based denoising.