.png){kind=icon}
.png){kind=icon}
Return
Weighted Patch-wise Minimum Pixel Prior for Blind Image Deblurring
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jfranklin.2026.108472.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Blind image deblurring aims to estimate a blur kernel from an intermediate latent image to recover the original sharp image. This paper proposes a novel image prior, called the Weighted Patch-wise Minimum Pixel prior (WPMP). Our approach leverages the ratio of the patch-wise minimum pixel (PMP) to the patch-wise maximum gradient (PMG), using the inverse of the PMG as the weight of the PMP. The WPMP prior considers both the intensity and gradient information of image patches, facilitating a more holistic approach to the deblurring process. Compared with the standalone priors, the WPMP enhances the contrast between clear and blurred images. Additionally, the non-overlapping patch design significantly boosts computational efficiency. Experimental results illustrate that our method outperforms existing methods, achieving an improvement of approximately 0.7dB in PSNR and 1.4% in SSIM. Our approach provides superior visual and quantitative performance on both natural and domain-specific images, underscoring the effectiveness of the proposed WPMP prior.
Keywords:
Blind image deblurring
Weighted Patch-wise Minimum Pixel prior
Image restoration
Patch-based prior
Deblurring algorithm
Journal
J
IF:
4.2
Papers:
822
Citations:
0