光学 精密工程, 2011, 19 (8): 1973, 网络出版: 2011-08-29
运动成像混合模糊的全变分图像复原
Total variation image restoration for mixed blur in moving image
运动模糊 离焦模糊 倒谱 全变分图像复原 梯度保真项 motion blur defocus blur cepstrum Total Variation (TV) image restoration gradient fidelity term
摘要
为了实现运动成像中视频帧内复原, 研究了帧内运动模糊和离焦模糊两种常见模糊类型带来的混合模糊问题。首先, 依据模糊图像的频谱特性定性判定模糊类型, 利用倒谱分析的方法定量估计模糊模型点扩散函数, 依据工程实际选取模糊参数进行仿真实验, 实现点扩散函数估计方法的准确性验证。然后, 采用耦合梯度保真项的改进型全变分图像复原算法, 约束点扩散函数估计误差对图像复原的影响, 并采用适应L1范数的Split-Bregman算法完成复原算法的数值实现。最后, 对算法性能进行验证, 完成仿真和实拍混合模糊图像的复原。实验结果表明, 倒谱分析估计点扩散函数的准确率达到90%。 复原算法能保持图像边缘和细节, 并有效抑制振铃效应, 帧内稳像的图像峰值信噪比为28.92 dB。
Abstract
In order to restore the image in the video frame of a moving image, the mixed blur that combines two common blur types, motion blur and defocus blur, was discussed. Firstly, blur types were determined according to the difference in spectrum features and the Point Spread Function (PSF) was estimated quantitatively using the cepstrum analysis. Blur parameters in simulation experiments were chosen based on engineering to verify the accuracy of the cepstrum analysis in the PSF estimation method. Then, a gradient fidelity term was coupled with total variation image restoration algorithm to constrain the impact of the PSF estimation error on image restoration and the Split-Bregman algorithm that is compatible with the L1 norm was adopted to accomplish the numerical computing in restoration algorithm. Finally, the simulation experiments and the real image restoration were carried to verify the performance of the algorithm. The results show that cepstrum analysis for estimating the PSF has an accuracy rate of 90%.The proposed algorithm can preserve edges and details, inhibit the ringing effect effectively and shows its Peak Signal to Noise Rate (PSNR) to be 28.92 dB.
石明珠, 许廷发, 张坤. 运动成像混合模糊的全变分图像复原[J]. 光学 精密工程, 2011, 19(8): 1973. SHI Ming-zhu, XU Ting-fa, ZHANG Kun. Total variation image restoration for mixed blur in moving image[J]. Optics and Precision Engineering, 2011, 19(8): 1973.