Figure 5.

Schematic diagrams of three denoising models. (a) The W5 architecture is composed of a skip connection and convolution layers. The convolution layers except for the last layer are followed by a ReLU activation function. (b) The DN architecture has a structure similar to W5. The difference is that each convolution layer has a skip connection to the output. (c) The N2N architecture is based on U-Net. The convolution layers except for the last layer are followed by a leaky ReLU activation function. The number of filters in the convolution layers of the encoder part is half of the number in the decoder part except for the last three layers.