Compressed sensing is widely applied in medical imaging, remote sensing, wireless communication, and other fields. However, deep learning-based compressed sensing relies on large amounts of training data, which is impractical in many applications. In this paper, we proposed a single shot image reconstruction method based on untrained neural network prior and Total Variation Regularization and Denoise Regularization. We demonstrate that this method not only reconstructs images but also improves reconstruction efficiency. Unlike various learning methods based on generative models, our method does not require pre-training on large datasets. Additionally, we employ l2 norm recovery, enabling the network to achieve more effective reconstruction results. We further introduce the regularization techniques we used, specifically Total Variation Regularization and Denoise Regularization, and demonstrated through experiments that they effectively reduce reconstruction errors. Additionally, experiments have shown that our proposed method effectively balances image reconstruction quality and processing time.