论文部分内容阅读
针对实时THz脉冲(T-ray)成像系统所成图像分辨率低、受1/f相关噪声干扰严重的特点,提出一种新的基于小波去噪的T-ray图像复原算法。对T-ray图像进行离散小波变换后,先利用广义交叉确认估计出各个分辨率层的噪声阈值,然后对每个分辨率层的高频子带进行迭代去噪,最后对去噪后的T-ray图像采用Jasson-Van-Cittert算法进行复原处理以提高分辨率。实验结果表明,该方法在提高T-ray图像分辨率的同时,能显著地抑制TH z成像系统的1/f相关噪声。创新之处在于以广义交叉确认作为T-ray图像中1/f噪声的估计方法,大幅度提高了图像信噪比(~5 dB),避免了噪声带来的复原算法中的不适定问题,达到较好的图像复原效果。
Aiming at the low resolution of the real-time THz pulse (T-ray) imaging system and the serious disturbance of 1 / f correlation noise, a new T-ray image restoration algorithm based on wavelet denoising is proposed. After discrete wavelet transform on T-ray image, the noise threshold of each resolution layer is estimated by generalized cross-validation, and then the high frequency sub-band of each resolution layer is iteratively denoised. Finally, the denoised T -ray image using the Jasson-Van-Cittert algorithm for recovery processing to improve resolution. Experimental results show that this method can significantly reduce the 1 / f correlation noise of THz imaging system while improving the resolution of T-ray images. The innovation is that broad cross-validation as a method of estimating 1 / f noise in T-ray images greatly improves the signal-to-noise ratio of the image (~ 5 dB) and avoids the ill-posed problem of the restoration algorithm caused by noise. Achieve better image recovery effect.