Proc. SPIE. 10697, Fourth Seminar on Novel Optoelectronic Detection Technology and Application
KEYWORDS: Signal to noise ratio, Holography, 3D image reconstruction, Digital holography, Image processing, Digital filtering, Denoising, Image quality, Image filtering, Optimal filtering, Terahertz detection
Terahertz digital holography, which can be used in nondestructive testing and other fields, is the effective combination of terahertz imaging and digital holography. The measurement accuracy may be influenced by the low energy of terahertz source and other factors; therefore it is very important to study denoising algorithms of terahertz digital holographic images. In this paper, optimally weighted bilateral filter (WBF) is applied in the preprocessing of guide image acquired by guided bilateral filter (GBF). This method can eliminate the noise in terahertz digital holography reconstructed image. We compare this algorithm with the original GBF with iterative reweighted least squares algorithm (IRLS). The experimental results show that the preprocessing method is better than the original algorithm.
In the process of recording terahertz digital hologram, the hologram is easy to be contaminated by speckle noise, which leads to lower resolution in imaging system and affects the reconstruction results seriously. Thus, the study of filtering algorithms applicable for de-speckling terahertz digital holography image has important practical values. In this paper, non-local means filtering and guided bilateral filtering were brought to process the real image reconstructed from continuous-wave terahertz coaxial digital hologram. For comparison, median filtering, bilateral filtering, and robust bilateral filtering, were introduced as conventional methods to denoise the real image. Then, all the denoising results were evaluated. The comparison indicates that the guided bilateral filter manifests the optimal denoising effect for the terahertz digital holography image, both significantly suppressing speckle noise, and effectively preserving the useful information on the reconstructed image.
Terahertz digital holography imaging technology is one of the hot topics in imaging domain, and it has drawn more and more public attention. Owing to the redundancy and translation invariance of the stationary wavelet transform, it has significant application in image denoising, and the threshold selection has a great influence on denoising. The denoising researches based on stationary wavelet transform are performed on the real terahertz image, with Bayesian estimation and Birge-Massart strategy applied to evaluate the threshold. The experimental results reveal that, Bayesian estimation combined with homomorphic stationary wavelet transform manifests the optimal denoising effect at 3 decomposition levels, which improves the signal-to-noise and preserves the image detail information simultaneously.
Terahertz (THz) holography is a frontier technology in terahertz imaging field. However, reconstructed images of holograms are inherently affected by speckle noise, on account of the coherent nature of light scattering. Stationary wavelet transform (SWT) is an effective tool in speckle noise removal. In this paper, two algorithms for despeckling SAR images are implemented to THz images based on SWT, which are threshold estimation and smoothing operation respectively. Denoised images are then quantitatively assessed by speckle index. Experimental results show that the stationary wavelet transform has superior denoising performance and image detail preservation to discrete wavelet transform. In terms of the threshold estimation, high levels of decomposing are needed for better denoising result. The smoothing operation combined with stationary wavelet transform manifests the optimal denoising effect at single decomposition level, with 5×5 average filtering.