Enhancing single-pixel imaging by improving one-dimensional signal through encoded image similarity

Single-pixel imaging (SPI) is a computational imaging technique with unique advantages. While recent research has advanced areas such as projection strategy optimization and image reconstruction algorithms, the processing of one-dimensional (1D) light intensity signals remains relatively underdeveloped. We propose a novel 1D signal processing method for SPI based on the similarity among encoded images, which utilizes the similarity among encoded images to perform weighted and linear combinations of 1D signals across different frames, effectively reducing noise and improving image quality. The proposed method demonstrates significant enhancements in image quality even under challenging conditions such as very low acquisition card sampling rates and low Signal-to-Noise Ratio (SNR). It is compatible with various existing reconstruction algorithms and, with adaptive modifications, can also be applied to other orthogonal projection bases, such as Fourier bases. Overall, our approach not only advances the processing of 1D signals in SPI but also provides a valuable strategy to address the challenges posed by insufficient sampling rates in future high-speed SPI. • A method enhances 1D signals in SPI through encoded image similarity. • Noise-robust performance under low SNR and low hardware sampling rates. • Compatible with existing algorithms and adaptable to various bases.