Temporal uniform sampling foveated Hadamard single-pixel imaging

Single-pixel imaging (SPI) is a computational imaging technique whose efficiency is fundamentally limited by the number of required measurements. Foveated imaging improves imaging efficiency by allocating a higher sampling density to regions of interest while compressing peripheral regions. Uniform-sampling foveated Fourier SPI overcomes the non-uniformity of conventional foveated sampling. However, its weighting-matrix-modulated Fourier patterns are grayscale and require binary modulation on digital micromirror devices (DMDs), introducing dithering errors. In this work, a temporal uniform-sampling foveated Hadamard SPI method is proposed. By exploiting the multi-frame measurement process of SPI, uniform sampling is achieved in the temporal domain rather than through spatial weighting. A Halton-based temporal sampling strategy is employed to provide approximately uniform temporal coverage while preserving binary Hadamard modulation. Consequently, efficient undersampled Hadamard sequences can be directly employed without additional weighting matrices or grayscale modulation, improving both imaging efficiency and reconstruction quality. Numerical simulations demonstrate that, under the same number of binary projections, the proposed method achieves higher global reconstruction quality compared with the dithered implementation of the existing uniform-sampling foveated SPI method. Experimental results using 4092 binary projections further validate the proposed method, showing an average PSNR improvement of 3.74 dB over the dithered spatial uniform-sampling foveated approach. The proposed temporal uniform-sampling framework provides a practical solution for efficient foveated single-pixel imaging and can be readily integrated into high-speed DMD-based imaging systems with precomputed measurement patterns.