Image And Video Deblurring Via Structural Prior

As camera,smart phones,and other mobile devices spread,people can take pictures and videos anytime and anywhere.However,motion blur is an artifact in photography caused by various sources,such as camera shake,moving objects,and depth variation in the relative motion between the camera and an imaged scene during exposure.These degraded images and videos lost most important information,so recovering these blurred images helps readers understanding scene content.In this paper,we focus on text image,natural image and video deblurring problems using different structural priors.The main contents and contributions of this work are summarized as follows:1)We study the problem of recovering the clear scene text image by proposing a text-specific multi-scale dictionaries based scene text image deblurring method.We exploit the text field characteristics and learn a series of text-specific multi-scale dictionaries and a natural scene dictionary for separately modeling the priors on the text and non-text fields.The text-specific dictionaries based text field reconstruction helps to deal with the different scales of strings in a blurry image effectively.2)We propose a novel low-rank prior for blind natural image deblurring.Our key observation is that directly applying a simple low-rank model to a blurry input image significantly reduces the blur even without using any kernel information,while preserving important edge information.The same model can be used to reduce blur in the gradient map of a blurry input.Based on these properties,we introduce an enhanced prior for image deblurring by combining the low rank prior of similar patches from both the blurry image and its gradient map.3)Video deblurring is a challenging problem as the blur is complex and usually caused by the combination of camera shakes,object motions,and depth variations.Therefore,traditional video deblurring methods based on uniform or global nonuniform cannot handle real-world videos.In this paper,we propose a novel video deblurring method to deal with complex blurs inherent in videos.We approximate pixelwise kernel with a quadratic function of bidirectional optical flows.The proposed blur model is based on the non-linear optical flow,which describes complex motion blur more effectively.In addition,we exploit semantic segmentation in each blurry frame to understand the scene contents and use different motion models for image regions to guide optical flow estimation,which further helps blur kernel estimation.