Wavelet-based threshold denoising for imaging hyperspectral data

Imaging spectroradiometer is highly susceptible to noise. Accurately quantitative processing with higher quality is obligatory before any derivative analysis, especially for precision agricultural application. Using the self-developed Pushbroom Imaging Spectrometer (PIS), a wavelet-based threshold (WT) denoising method was proposed for the PIS imaging hyperspectral data. The WT with PIS was evaluated by comparing with other popular denoising methods in pixel scale and in regional scale. Furthermore, WT was validated by chlorophyll concentration retrieval based on red-edge position extraction. The result indicated that the determination coefficient R2 of the chlorophyll concentration inversion model of winter wheat leaves was improved from 0.586 to 0.811. It showed that the developed denoising method allowed effective denoising while maintaining image quality, and presented significant advantages over conventional methods.