Multi-scale Progressive Fusion Network

Introduction: Multi-scale Progressive Fusion Network (MSFPN) is a neural network representation designed for single image deraining, which helps remove the rain streaks from images. The network aims to leverage the related information available on different scales of rain streaks to improve the derain performance. Deraining using MSFPN: With MSFPN, we use the Gaussian kernel to down-sample the original image to generate the Gaussian pyramid rain images. This image is then fed into the Multi-

1 / 1