Research On Terahertz Time Domain Spectral Imaging

Terahertz time-domain spectroscopy(THz-TDS)imaging technology can effectively achieve millimeter or sub-millimeter order transmission imaging,and solve the subwavelength imaging distance problem of THz-TDS near-field high-resolution imaging technology and reduce the imaging system complexity.For the low spatial resolution of THzTDS imaging limited by the diffraction limit theory,experimental noise,terahertz source power jitter,and medium inhomogeneity in the imaging process lead to serious image noise,low contrast,and blurred image details.This paper studied image denoising,enhancement and single-image resolution reconstruction algorithms in the THz-TDS imaging,which achieves noise removal and enhanced image detail.Firstly,the THz-TDS imaging system transmission variables(spectrum,beam divergence,depth of focus,and absorption coefficient)are incorporated into the Gaussian beam distribution to establish the point spread function(PSF)model of the THz-TDS imaging system,establishing terahertz imaging function by convolution of PSF and measured target function,and the imaging system noise and imaging quality are simulated.Simulation experiments show that the simulated terahertz image shows a high accuracy.Secondly,for the problem of terahertz image noise caused by terahertz source power jitter and medium inhomogeneity,a terahertz image denoising algorithm based on adaptive bilateral filtering is proposed to implement edge information preservation and noise filtering.Increased peak signal to noise ratio(PSNR).At the same time,the enhancement of the image detail is not perfect after the enhancement of the wavelet-domain non-linear processing combined with the multi-scale Retinex(DWT_MSR)THz image enhancement algorithm.The multi-scale detail enhancement method is used to improve the DWT_MSR and enhance the contrast of the enhanced image.Detailed information improves visual information assurance.Finally,the high-resolution reconstruction of low-resolution terahertz images is realized by the image super-resolution reconstruction method with sparse representation,non-local self-similarity and gradient regularization constraints.A single image super-resolution(SISR)reconstruction model was proposed by the gradient prior regularization constraint of the nonlocal self-similarity regularization constraint and joint optimization(JOR)model.The SISR model was multi-variable optimized and fuzzy was implemented.Low-resolution terahertz images have good high-resolution reconstruction,and the reconstructed high-resolution images have sharper edges and details.