An improved asymptotic denoising algorithm for infrared images of electrical equipment based on nonlocal means

Infrared images are susceptible to Gaussian noise during acquisition and transmission. To meet the quality requirements for infrared image data in image processing, an asymptotic Non-Local Means (NLM) algorithm based on an improved Scharr operator was proposed for denoising infrared images of electrical equipment. Firstly, the RGB infrared images were converted to the Y channel. Secondly, the improved Scharr operator was used to adjust the computation weights of the NLM algorithm, better preserving the edge details of the images. To optimize the denoising effect, a multi-layer pyramid was constructed, and a layer-by-layer denoising fusion method was adopted to obtain the initially denoised images. Finally, for the residual noise in the initially denoised images, asymptotic denoising was employed, adjusting the filter parameters based on the image characteristics for secondary denoising, followed by image reconstruction to obtain the denoised images. Experimental results showed that the proposed algorithm achieved clear and complete denoised images, with Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM) improved by 0.78 to 4.95 and 0.01 to 0.08, respectively, compared to other methods. This indicated that the method is effective for high-efficiency denoising of infrared images of electrical equipment.