Abstract
Machine learning based methods for blind deblurring are efficient to handle real-world blurred images, whose blur may be caused by various combined distortions. However, existing multi-level architectures fail to fit images of various scenarios. In this paper, we propose a scale-iterative upscaling network (SIUN) that restores sharp images in an iterative manner. It is not only able to preserve the advantages of weights sharing across scales but also more flexible when training and predicting with different iterations to fit different images. Specifically, we bring in the super-resolution structure instead of the upsampling layer between two consecutive scales to restore a detailed image. Besides, we explore different curriculum learning strategies for both training and prediction of the network and introduce a widely applicable strategy to make SIUN compatible with different scenarios, including text and face. Experimental results on both benchmark datasets and real blurred images show that our method can produce better results than state-of-the-art methods. Code is available at https://github.com/minyuanye/SIUN.
Highlights
Image deblurring, aiming to recover a sharp image from its blurred source, has long been a challenging and fundamental problem in computer vision and image processing
We propose a scale-iterative upscaling network (SIUN)
There are several advantages of this new framework: (1) Like scale-recurrent structure, it has far fewer trainable parameters through weights sharing than scale-cascaded structure; (2) Compared with previous fixed-level architecture, it is more flexible when training and predicting with variable iterations to fit different images; (3) Instead of the upsampling layer used in [10] and [11], it adapts residual dense network (RDN) [13], a super-resolution architecture for upscaling so that more details of the image can be restored; (4) It is more compatible with diverse scenarios of real-world images, including text and face, with the curriculum learning strategy designed for both training and prediction
Summary
Image deblurring, aiming to recover a sharp image from its blurred source, has long been a challenging and fundamental problem in computer vision and image processing. INDEX TERMS Blind deblurring, curriculum learning, scale-iterative, upscaling network. There are several advantages of this new framework: (1) Like scale-recurrent structure, it has far fewer trainable parameters through weights sharing than scale-cascaded structure; (2) Compared with previous fixed-level architecture, it is more flexible when training and predicting with variable iterations to fit different images; (3) Instead of the upsampling layer used in [10] and [11], it adapts residual dense network (RDN) [13], a super-resolution architecture for upscaling so that more details of the image can be restored; (4) It is more compatible with diverse scenarios of real-world images, including text and face, with the curriculum learning strategy designed for both training and prediction.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call