Missing Data Estimation by Separable Deblurring

Today's technology allows butting a few sensor arrays to a high precision in order to capture a twodimensional image of large area. The most serious defect caused by this butting technique is the gap between sub-arrays. This paper proposes an image restoration method to recover the missing data using the information of blur. We claim that by making a reasonable assumption that the blur in real world is usually Gaussian blur, we can take advantage of the separability property of Gaussian kernel to separate the deblurring process, and recover the missing data during the separated deblurring. We also prove that the problem is well-conditioned, and the algorithm we used is backward-stable. Experimental results are provided.