Noise Estimation and Type Identification in Natural Scene and Medical Images using Deep Learning Approaches

Table 3

Training parameters for CNN model for both medical images and natural scene images.

Parameters	Descriptions
Input layer	112 × 112 × 1
First layer	Filter kernel size is 3 × 3 with 8 filters with padding.
Batch normalization and ReLU layer
Second layer	Kernel size is 3 × 3 with 16 filters with padding
Batch normalization and ReLU layer
Third layer	Kernel size is 3 × 3 with 32 filters with padding
Batch normalization and ReLU layer
Fourth layer	Kernel size is 3 × 3 with 32 filters with padding
Softmax	Fitting tool
Learning rate	1 × 10^−4
Validation frequency	30