Improvement of terahertz images by adaptive discrete cosine transform (DCT)-based denoising

Due to certain hardware limitations the quality of terahertz images is often lower than desired, which makes it difficult to extract valuable information from them. The goal of this paper is to investigate possibilities to overcome some of these limitations by means of digital image processing. The research is held on a set of images obtained at different distances from the source of terahertz radiation at 0.1 THz frequency. It is shown that the noise in these images is mixed and has a significant level of spatial correlation. For image quality enhancement a fully automatic denoising method based on the use of a discrete cosine transform with a spatially adapted spectrum is proposed. It is shown that despite an initially low spatial resolution of terahertz images and intensive noise, it provides a good noise reduction with a good preservation of edges, which allows one to noticeably improve the quality of these images and make them more convenient for visual analysis carried out by a human operator.