摘要
正则化技术为基础,提出自适应的TV(Total Variation)模型,实现平滑去噪和保护图像的纹理细节,将研究图像复原转化为求解代价函数的最小化过程,并运用共轭梯度法搜索函数极值,以实现湍流退化图像的复原过程。实验结果表明,本文所提出的算法能够在较大程度上改善图像的质量,自适应全变分技术和正则化技术的合理使用能有效地保护图像边缘和去噪,并具有较好的收敛速度和稳定性。本文提出的算法计算强度适中,适合应用于具有较高实时性要求的航空航天、军事及农业等领域。
Based on the regularization technique, adaptive TV (Total Variation) model was put forward to denoise smoothly and protect image texture details. The image restoration study was converted into solving the cost function minimization process, and conjugate gradient method was used to search function extremum in order to realize the turbulence degraded image restoration process. The presented algorithm could improve the quality of the image in a larger extent, the rational use of adaptive total variation and regularization technique can effectively protect image edge and denoise, and has better convergence speed and stability. It calculates at moderate strength, and is suitable for application in areas which has high real-time requirements like aerospace, military, agriculture industry, etc.
出处
《中国农机化学报》
北大核心
2014年第4期257-259,273,共4页
Journal of Chinese Agricultural Mechanization
关键词
图像复原
点扩散函数
TV模型
自适应总变分模型
共轭梯度法
image restoration
point spread function
TV model
adaptive total variation model
conjugate gradient method