Over the past decades, millimeter wave and terahertz radiation has received a lot of interest due to advances in
emission and detection technologies which allowed the widely application of the millimeter wave and terahertz imaging
technology. This paper focuses on solving the problem of this sort of images existing stripe noise, block effect and other
interfered information. A new kind of nonlocal average method is put forward. Suitable level Gaussian noise is added to
resonate with the image. Adaptive color block-matching 3D filtering is used to denoise. Experimental results
demonstrate that it improves the visual effect and removes interference at the same time, making the analysis of the
image and target detection more easily.