Image denoising algorithm based on wavelet transform and convolutional neural network

In the process of image generation and sensing, the image is often disturbed by the noise, which will increase the difficulty to the subsequent image processing and even bring the huge economic losses to some production activities. Combined the advantages of stationary wavelet transform with convolutional neural network, an effective image denoising algorithm was proposed. In the training stage, the proposed algorithm was adapted to decompose the image with a stationary wavelet of scale 1, and the high and low frequency components were input into the four designed residual networks for training respectively. In the test phase, the inverse wavelet transform was used to obtain the final predicted image. The experimental results show that when the level of Gaussian white noise reaches σ = 50, the mean of peak signal to noise ratio (PSNR) and structural similarity index method (SSIM) of the denoised image can reach 28.37 dB and 0.808 0. This algorithm can effectively remove the Gauss white noise and natural noise in visible image, and the noise generated by remote sensing image in the sensing process. Moreover, the proposed algorithm can preserve the edge and texture details of the image while removing the image noise.