Development of Weak Signal Recognition and an Extraction Algorithm for Raman Imaging

Improving time resolution of Raman imaging is essential for the observation of dynamic processes involved in interfacial catalysis and biological systems. The crucial step is how to recognize and extract weak Raman signals overwhelmed in the strong noise under the low signal-to-noise ratio (SNR) condition. Here, by exploring the relationship between the SNR of a single Raman spectrum and the structural similarity (SSIM; the key parameter evaluating the image quality) of the whole image, we determined a semiempirical threshold with SNR = 0 dB for clear imaging for the first time. Therefore, we proposed one signal processing algorithm for fast Raman imaging by reconstructing the Raman spectrum with the aid of weak signal processing: extracting the reliable Raman signal of the target under the low SNR and then determining the suitable scanning time to obtain the Raman image with a trustworthy image quality. In the first step, fast Fourier transform (FFT), least squares, and 2-D median filter are sequentially applied to improve the SNR of each raw Raman spectrum. In the second step, a local SNR evaluation strategy is developed to predict image quality as well as the determination of clear imaging. The proposed method was successfully applied to the fast imaging of the cell under the low SNR condition.