.png){kind=icon}
.png){kind=icon}
Return
Image restoration based on integrated sparse representation and self-representation learning
.png){kind=icon}
PRE
AI
Y
H
C
G
DOI：10.1117/1.JEI.35.2.023037.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Traditional image restoration methods often rely on sparse representation and low-rank modeling, which excel at capturing local structures but may overlook nonlocal similarities between image blocks and then lead to texture detail loss and global structure degradation. To address these limitations, we propose a innovative image restoration model that integrates sparse representation with self-representation learning. By leveraging the internal structure of images, each block is represented as a linear combination of other blocks, preserving details and suppressing noise. Sparse representation is combined with self-representation learning within the maximum a posteriori probability framework to derive an optimization problem, which is iteratively solved using the alternating minimization algorithm. Experimental results show that our model demonstrates good performance in the fields of image denoising and inpainting, especially outperforming other traditional methods in terms of maintaining texture details and global consistency.
Keywords:
image denoising
image restoration
image inpainting
sparse representation
self-representation learning
low rank
Journal
J
IF:
1
Papers:
109
Citations:
2.7K
Organization
Cited Papers