Deep Signal-Dependent Denoising Noise Algorithm

Although many existing noise parameter estimations of image signal-dependent have certain denoising effects, most methods are not ideal. There some problems with these methods, such as poor suppression smooth details, lack flexible ability, etc. To solve problems, in this study, we propose a deep algorithm. The algorithm combines the model method convolutional neural network. We use level and together input network to obtain wider range levels than single input. In network, features extracted by multi-layer residuals, which solves difficult problem training. Extensive experiments demonstrate that our estimation has good performance.