Separable Hadamard single-pixel imaging for high-resolution reconstruction.

Area of Science:

• Computational Imaging
• Optical Engineering
• Signal Processing

Background:

• Hadamard single-pixel imaging (HSI) is a promising computational imaging technique.
• Existing HSI methods face challenges with high data throughput and computational demands, hindering high-resolution image reconstruction.

Purpose of the Study:

• To address the limitations of conventional HSI, we introduce a separable HSI (SHSI) model.
• The goal is to significantly reduce storage requirements and accelerate imaging speed for efficient high-resolution reconstruction.

Main Methods:

• Investigated the transformation between Hadamard spectrum distribution and bidirectional compressed sensing.
• Proved that 2D sampling masks can be decomposed into 1D vectors from separable measurement matrices.
• Developed an iterative alternating optimization algorithm for optimal separable sampling strategy, decomposing 2D masks into separable 1D vectors.

Main Results:

• The SHSI model drastically reduces storage and accelerates imaging speed.
• Real-time pattern generation with ultra-low storage and computational overhead was achieved.
• High-efficiency and high-resolution (1024x1024) image reconstruction was demonstrated through simulations and experiments.

Conclusions:

• The proposed SHSI model offers a novel strategy for efficient high-resolution image reconstruction.
• SHSI overcomes the bottlenecks of data throughput and computational time in traditional HSI.
• This technique enables practical, high-performance single-pixel imaging applications.