Extracting phase from a single-frame fringe pattern is a key and challenging problem in fringe projection 3D measurement, especially for dynamic 3D measurement. We propose a single-shot phase extraction approach based on a low-pass filter with a well-trained image denoiser. Various comparative experiments have verified the effectiveness of the proposed method. More importantly, we associate the single-shot phase extraction problem with image denoising using deep learning. Thus more existing well-designed deep neural network models can be reused in the proposed method, without having to design a new model.