Multi-exposure detail block fusion algorithm for polarization-preserving denoising of low-light polarization spectral...

A noise reduction algorithm is proposed for polarization spectral images under low-light conditions. The algorithm combines multi-exposure image fusion with the preservation of polarization features through a collaborative processing framework in both frequency and spatial domains. It utilizes the discrete cosine transform (DCT) to reduce high-frequency noise and block matching techniques to preserve polarization specifics, thereby improving accuracy and reducing noise in the degree of linear polarization (DoLP) and angle of polarization (AoP) data. This method proves advantageous for reducing noise and conserving polarization in low-light imaging scenarios such as intelligent transportation and visual surveillance.