A Novel Passive Millimeter Wave Image Noise Suppression Method Based on Pixel Non-local Self-similarity

Abstract

To solve the problem of mixed noise in a passive millimeter-wave (PMMW) imaging system that affects object detection, recognition, and classification, this paper proposes a blind denoising algorithm based on pixel non-local self-similarity (PNSS) prior to PMMW images. Firstly, an adaptive filtering algorithm is introduced, utilizing PNSS prior to estimating the noise intensity and improving the problem of noise residual caused by parameter uncertainty in traditional filtering processes. Secondly, a three-level joint denoising algorithm is developed, accompanied by an iterative regression algorithm to effectively filter the mixed noise in PMMW images while preserving image contours. Finally, the effectiveness of the proposed method is demonstrated through a comparison with patch similarity-based prior denoising methods and high-dimensional mixed noise denoising methods. Experimental results substantiate that the proposed PNSS blind denoising method successfully suppresses mixed noise in PMMW images, enhances subjective visual perception, and presents a novel approach for denoising under various PMMW imaging mechanisms.