Non-uniform motion deblurring with blurry component divided guidance

Blind image deblurring is a fundamental and challenging computer vision problem, which aims to recover both the blur kernel latent sharp from only blurry observation. Despite superiority of deep learning methods in have displayed, there still exists major challenge with various non-uniform motion blur. Previous simply take all features as input decoder, handles different degrees (e.g. large blur, small blur) simultaneously, leading challenges for generation. To tackle above problems, we present two-branch network deal images via component divided module, divides an into two components based on representation degree. Specifically, attentive blocks are employed learn attention maps exploit useful feature representations regions. Then, blur-aware fed reconstruction decoders respectively. In addition, new fusion mechanism, orientation-based fusion, proposed merge branches. Both qualitative quantitative experimental results show that our method performs favorably against state-of-the-art approaches.