Eliminating Dual-Frame Redundancy in Binary DCT Single-Pixel Imaging via Zero-Threshold Modulation

Discrete cosine transform-based single-pixel imaging (DCT-SPI) enables high-quality sub-Nyquist imaging owing to its strong energy compaction properties. However, its reliance on grayscale modulation is incompatible with binary high-speed digital micromirror devices (DMDs), and existing binary adaptations introduce dual-frame redundancy, severely limiting imaging speed. Here, we propose a two-dimensional compressive binary DCT-SPI framework. By applying a zero-threshold binarization strategy to the discrete cosine patterns, binary illumination patterns are generated that completely eliminate dual-frame redundancy. This single-pattern approach fully exploits the high-speed switching capability of DMDs, approaching the theoretical limit of modulation efficiency. Furthermore, by exploiting the separability of the discrete cosine basis, we reformulate the measurement model into a matrix form and reconstruct it using a two-dimensional compressed sensing strategy, which significantly reduces the computational burden. Experimental results demonstrate that the proposed method achieves high-fidelity reconstruction at a resolution of 256×256 with only 1936 measurements, while maintaining robustness to noise.