Wide-field, native-resolution single-pixel imaging via Kronecker product modeling and fractional-order total variation regularization

Conventional single-pixel imaging (SPI) is fundamentally constrained by the heavy computational and memory overheads imposed by large-scale inverse problems, which restrict the achievable field of view (FOV) far below the physical capacity of spatial modulation devices and hinder practical high-throughput imaging. To address this limitation, we propose a wide-field, native-resolution SPI framework termed Kronecker product-based fractional-order total variation SPI (KronFOTV-SPI). This framework integrates Kronecker product modeling with fractional order total variation (FOTV) regularization. By factorizing the large measurement matrix into two compact submatrices, the proposed method fully utilizes the native pixels of the modulation device while greatly reducing computational and memory costs. FOTV further improves reconstruction fidelity by preserving fine details across wide-field regions and suppressing noise and artifacts. Experimental results demonstrate that the proposed KronFOTV-SPI method achieves full native resolution reconstruction at 1 0 2 4 × 7 6 8 pixels across the entire imaging region, delivering a 4 8 × larger FOV than traditional small FOV SPI, along with 1.8 × and 2.2 × expansions over parallel SPI and pixel binning SPI. High reconstruction fidelity is maintained without spatial averaging artifacts, while memory consumption is reduced to approximately 12.5 MiB, and reconstruction is completed within approximately 6.25 s for a 700k-pixel image. These results demonstrate that the proposed KronFOTV-SPI framework provides a scalable and computationally efficient solution for wide-field native-resolution SPI, with significant potential for practical large-scale imaging applications.