Density-aware Single Image De-raining using a Multi-stream Dense Network

Abstract

Single image rain streak removal is an extremely challenging problem due to the presence of non-uniform rain densities in images. We present a novel density-aware multi-stream densely connected convolutional neural network-based algorithm, called DID-MDN, for joint rain density estimation and de-raining. The proposed method enables the network itself to automatically determine the rain-density information and then efficiently remove the corresponding rain-streaks guided by the estimated rain-density label. To better characterize rain-streaks with different scales and shapes, a multi-stream densely connected de-raining network is proposed which efficiently leverages features from different scales. Furthermore, a new dataset containing images with rain-density labels is created and used to train the proposed density-aware network. Extensive experiments on synthetic and real datasets demonstrate that the proposed method achieves significant improvements over the recent state-of-the-art methods. In addition, an ablation study is performed to demonstrate the improvements obtained by different modules in the proposed method. Code can be found at: https://github.com/hezhangsprinter

Accepted in CVPR'18

References in corpus (7)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Depth Map Prediction from a Single Image using a Multi-Scale Deep Network
• Multi-style Generative Network for Real-time Transfer
• Hidden Two-Stream Convolutional Networks for Action Recognition
• MDNet: A Semantically and Visually Interpretable Medical Image Diagnosis Network
• Joint Transmission Map Estimation and Dehazing using Deep Networks
• A Recurrent Encoder-Decoder Network for Sequential Face Alignment

Cited by in corpus (13)

• Gated Context Aggregation Network for Image Dehazing and Deraining
• Residual-Guide Feature Fusion Network for Single Image Deraining
• NASNet: A Neuron Attention Stage-by-Stage Net for Single Image Deraining
• Non-locally Enhanced Encoder-Decoder Network for Single Image De-raining
• Two Stream Networks for Self-Supervised Ego-Motion Estimation
• Syn2Real Transfer Learning for Image Deraining using Gaussian Processes
• M2GAN: A Multi-Stage Self-Attention Network for Image Rain Removal on Autonomous Vehicles
• Rain Removal By Image Quasi-Sparsity Priors
• Removing rain streaks by a linear model
• Model Watermarking for Image Processing Networks
• Residual Squeeze-and-Excitation Network for Fast Image Deraining
• Exposure Interpolation by Combining Model-driven and Data-driven Methods
• A De-raining semantic segmentation network for real-time foreground segmentation