|
|
 |
| |
|
|
|
Optimizing Conjugate Gradient Methods for Image Recovery from Salt and Pepper Noise |
|
|
|
PP: 627-636 |
|
|
|
doi:10.18576/jsap/140610
|
|
|
|
Author(s) |
|
|
|
Basim A. Hassan,
Talal Alharbi,
|
|
|
|
Abstract |
|
|
| This study focuses on image restoration from salt-and-pepper impulse noise using an enhanced Conjugate Gradient (CG) method. The proposed approach adopts a two-phase process: first, an adaptive median filter identifies corrupted pixels; second, a smooth optimization problem replaces a non-smooth one to reconstruct the image. New formulas derived from the Taylor series are introduced to define the BBY algorithms, which improve convergence and maintain essential image features like edges. The convergence analysis demonstrates that the proposed methods satisfy descent conditions and are globally convergent. Experimental results using standard images (e.g., Lena, House, Elaine, Cameraman) show that BBY algorithms outperform the classical FR method in both computational efficiency and image quality, measured using PSNR. Numerical comparisons reveal lower iteration counts and higher PSNR values for BBY. Theoretical assumptions are validated with practical experiments. The BBY framework proves robust and effective for real-world noisy image restoration. Overall, the study demonstrates the strength of combining optimization theory with practical image processing techniques. |
|
|
|
|
|
|
|
