TY - JOUR
AU - N, Sharmila
AB - Deblurring an image has been a long researched problem. This problem is very complex due to the lack of sufficient information about the blur parameters. Image deblurring is important in applications such as remote sensing where the same image scene comprising of moving objects cannot be captured again. Since the captured image is the only known quantity, from which the blur parameters affecting it and the sharp image has to be estimated it is an illposed problem. In this paper we present a novel image deblurring algorithm which makes use of region specific priors and techniques for image deblurring. This is based on the idea that different image regions require different techniques for effective image deblurring. Applying the same technique to deblur the entire image results in a deblurred image which is sharp in only some regions whereas some regions are not effectively deblurred or have ringing artifacts. The proposed method makes use of a l1 relaxed l0 prior on the sharp edges of the image to effectively enhance the true edges of the image. An l1/l2 norm on the low peaks to avoid blurring while retaining the true values in these regions and lp norm prior on the uniform regions to avoid the generation of ringing artifacts. On an average there is about 18% increase in PSNR values and a 5% increase in the SSIM values in comparison with the existing state of the art methods.
TI - Image region driven prior selection for image deblurring
JF - Multimedia Tools and Applications
DO - 10.1007/s11042-023-14335-y
DA - 2023-07-01
UR - https://www.deepdyve.com/lp/springer-journals/image-region-driven-prior-selection-for-image-deblurring-u96PkSDhvd
SP - 24181
EP - 24202
VL - 82
IS - 16
DP - DeepDyve
ER -