MSDU-Net: A Multi-Scale Dilated U-Net for Blur Detection

A blur detection problem which aims to separate the blurred and clear regions of an image is widely used in many important computer vision tasks such object detection, semantic segmentation, face recognition, attracting increasing attention from researchers industry recent years. To improve quality separation, have spent enormous efforts on extracting features various scales images. However, matter how extract fuse these synchronously still a big challenge. In this paper, we regard as segmentation problem. Inspired by success U-net architecture for propose multi-scale dilated convolutional neural network called MSDU-net. model, design group feature extractors with convolutions textual information at different same time. The U-shape MSDU-net can different-scale texture generated support task. We conduct extensive experiments two classic public benchmark datasets show that outperforms other state-of-the-art approaches.