<p>In recent years, the plug-and-play alternate direction method of multipliers (PnP-ADMM) has been extensively studied, leading to significant advancements in image restoration. However, existing PnP-ADMM algorithms typically assume the noise level parameter in the denoiser to be a known constant. This assumption limits the adaptability of the denoiser to varying noise levels and may compromise the algorithm’s convergence due to its dependence on the noise level. To address these limitations, this paper establishes a relationship between the noise level and the denoising intensity of the denoiser, enabling the denoiser in the PnP-ADMM algorithm to adaptively adjust its denoising intensity. Consequently, we propose a PnP-ADMM algorithm with embedded noise level estimation, termed the embedded noise level estimation PnP-ADMM (NLPnP) algorithm. We also prove that the iterative sequence generated by the algorithm is a Cauchy sequence, thereby guaranteeing its convergence. Additionally, experiments on grayscale images, color images, and various image datasets demonstrate the effectiveness of the algorithm in image restoration tasks, such as image deblurring, super-resolution, and image inpainting. Numerically, the NLPnP algorithm achieves higher peak signal-to-noise ratio and structural similarity values compared to the PnP algorithm.</p>