Abstract: Aiming at the drawbacks of the denoised algrithms, that can only removal one or two specifickind of noise and are invalid for others, we combine some excellent neural network model and proposed the image mixed noise removal algorithm based on convolutional neural network. 9 convolution layers are adopted, and noise images are trained through feature extraction, shrinking, non-liner mapping, expanding and reconstruction. Experimental results show that the algorithm achieves better denoised results and is suitable for different kinds of mixed noise images. The subjective visual effect and objective evaluation indices are improved obviously.