Speckle noise will inevitably exist in medical ultrasound images. Therefore, this study proposes a medical ultrasound image noise reduction model based on the dual attention mechanism to effectively remove noise in the medical ultrasound images. First, due to the limited number of medical ultrasound images, we rotated and zoomed 400 images in Berkeley dataset to receive 23700 images. Then, we added speckle noises using the speckle noise model to simulate the ultrasonic images. Second, during the construction process of the noise reduction model, aiming at some disadvantages of traditional convolutional neural networks in the feature extraction, we introduced position attention mechanism, channel attention mechanism, and full convolutional network to improve the existing model and build a better ultrasonic image noise reduction model. Finally, we introduced a batch normalization operation to prevent the gradient from disappearing during the model training process. The experimental results show that the noise suppression effects of 11 simulated ultrasound images and 2 real ultrasound images (the physical body membrane and liver ultrasonic images) are better than other models in terms of visual observation effect and objective evaluation index. Therefore, the proposed model is an effective noise suppression model for medical ultrasound images. It can effectively reduce speckle noise and retain the image details.