High-speed motion image deblurring using referenceless image quality assessment

摘要

A new method is developed to deblur seriously blurred images taken in a long exposure of camera in the fast motion. It is different from most of the existing deblurring methods that result in a poor quality for processing serious blur images, and few papers concerned the deblurred image quality, the authors study referenceless image quality assessment method to deblurring image. A blurred image is modelled based on a fast motion scene. Due to the consistent characteristics of direction motion blurs and ringing artefacts generated with the direct deconvolution, an adjacent-column-greyscale-difference index is proposed, which combines with an effective spatial quality evaluator to reduce the artefacts and improve quality of the deconvolution image. The process of the proposed method is automatic and fast in the image deblurring, it is verified on both the synthetic-blurred and the real-scene-blurred images, which outperforms the existing methods in reducing serious blurs of forward motion images.