About training

What's the training data seem like? Are the input generated by just adding noise to the clean images?
Thanks for your reply ：)
---
训练数据是否仅仅通过在干净的图片上添加高斯噪声生成？要通过模糊核模糊吗？