Removing motion blur in images
18 July 2014. NUS professors in Department of Mathematics have developed new mathematical method to correct image blurring caused by camera shake.
Despite all the recent advances in digital photography, motion blur is one of the most annoying effects that often creep into your photos taken with a hand-held camera. Is there a way to remove this blurring without knowing beforehand exactly how the camera was shaken? Conceptually, removal of motion blurring from the resultant image can be formulated as a blind de-convolution problem. However blind de-convolution is a very challenging ill-posed non-linear inverse problem. There have been only few attempts to tackle this problem in the past. Now, building upon the sparse approximation of images by wavelet frame theory, the research team led by Professor JI Hui has developed novel mathematical models and computational methods to handle blind image de-convolution. These new techniques have led to some powerful software toolboxes that can be used to routinely remove motion blur in photos (see Figure). These results and computational tools are also applicable to many other non-linear inverse problems arising from imaging science. This work has been published in the IEEE Transactions on Image Processing.
A demonstration of removing motion blurring from a single photo.
(Left) Input motion-blurred photo (Right) Output sharp photo (Image credit: JI Hui)
Cai J, Ji H, Liu C, Shen Z. "Framelet based blind image deblurring from a single image." IEEE Transactions on Image Processing 21 (2012) 562.