Research on image fusion noise reduction method based on AFM scanned cell image in tapping

Scanning cells with Tapping-Mode Atomic Force Microscopy (TM-AFM) often introduces significant noise in the central region of the topography image, whereas the corresponding amplitude image remains relatively clear. Leveraging the strong correlation between these two image types, this study proposes a noise reduction method for cellular topography images through the fusion of amplitude and topography data. The efficacy of this approach was evaluated using SMCC-7721 cells exhibiting four distinct morphologies: triangular, star-shaped, spindle, and nuclear. Four fusion algorithms—weighted average, PCA, dual-channel PCNN, and a six-layer Laplacian pyramid—were applied and assessed. Subjective evaluation indicated that the weighted average and six-layer Laplacian pyramid algorithms yielded superior visual outcomes, although the former led to information loss in central cell regions. Objectively, quantitative metrics (EN and g) confirmed that all fused images scored higher than original topography images. The six-layer Laplacian pyramid algorithm achieved top-tier performance in both subjective and objective assessments, effectively suppressing noise while preserving structural details. Thus, it is identified as the optimal strategy for enhancing TM-AFM image quality, providing researchers with a robust and efficient noise-reduction tool for cellular imaging.