Single-Pixel Imaging Via Optimizing The Measurement Matrix

Single-pixel imaging is a new computational imaging approach that shares similarities with ghost imaging.It allows for the acquisition of spatial information of an object using spatially non-resolvable photodetectors and reconstructs an image through computation.Single-pixel imaging does not require complex imaging devices such as imaging sensors or cameras,making it a more cost-effective and widely applicable imaging solution.Due to its miniaturization and wide spectral range,single-pixel detectors are suitable for various imaging schemes across different spectral regions.In addition,single-pixel detectors have higher quantum efficiency and filling factor than imaging sensors.Compressive sensing-based single-pixel computational imaging allows for the acquisition of spatial information using spatially non-resolvable photodetectors and image reconstruction through computation.However,single-pixel imaging is sensitive to noise and interference,requiring higher signal-to-noise ratios and better environmental conditions to ensure imaging accuracy.During the sampling process,various noise contaminations such as detector noise,instantaneous random fluctuations and long-term drift of illumination,spatial non-uniformity of illumination,as well as factors such as the reflectivity and transmittance of the medium,can decrease the signal-to-noise ratio of reconstructed images,affecting the further practical application of single-pixel technology.Based on single-pixel imaging,we propose an optimized compressed sensing single-pixel imaging scheme for measuring matrices.This scheme uses the total variation augmented lagrangian alternating direction algorithm(TVAL3)that has better image recovery performance in compressed sensing as the image recovery algorithm.The optimized measurement matrix is introduced into the image recovery calculation of TVAL3 to replace the original measurement matrix.The new measurement matrix exhibits a diagonal sparse state and has lower pathological degree and computational complexity during the compressed sensing computation.The optimized measurement matrix enables the recovered image to have strong optimization ability against the noise pollution problem mentioned above.Compared with the imaging results before improvement,the sampling rate is reduced while the noise interference is significantly eliminated,and the object spatial intensity information is presented more clearly.In this paper,we also explored experimental research using a liquid crystal phase spatial light modulator to generate non-rayleigh nondiffracting speckles as ghost imaging light sources.Meanwhile,we introduced the progress of customizing speckles using spatial light modulators and selected Raleigh speckles,different contrast super-Raleigh speckles,nondiffracting speckles,and non-rayleigh nondiffracting speckles with different contrasts for comparative experiments.The experimental results show that,under the same speckle size conditions,whether it is correlation imaging or compressed sensing imaging,analyzing the imaging results under different contrast conditions for different speckles and comparing the signal-tonoise ratio and mean square error,it can be concluded that non-rayleigh nondiffracting speckles as the light source have better imaging effects than other selected light sources,and the object information is more prominent in the background.