In image processing, denoising can provide high-quality data for subsequent processing. Current denoising methods can be categorized into three types: learningbased methods, model-based methods, and filtering-based methods. Although the first two methods excel at handling complex noise, traditional filtering methods remain advantageous in scenarios requiring high real-time performance and limited computational resources.  In this paper, an adaptive fuzzy nonlinear filtering algorithm (NCFR algorithm) based on neighborhood-controlled fuzzy rough set model is proposed to effectively remove salt-and-pepper noise from images. The proposed NCFR algorithm introduces fuzzy membership functions to address the uncertainties and fuzziness between image pixels, and utilizes the noise-free pixels within an adaptive neighborhood to restore noisy pixels, thereby efficiently reducing noise while preserving image details.  Experimental results demonstrate that the NCFR algorithm significantly improves image restoration quality under various noise densities. In particular, at high noise densities exceeding 95%, the algorithm shows superior performance in terms of PSNR and SSIM indices, better preserving structural characteristics and texture details of the images.