具有噪声估计机制的偏差迭代图像复原

研究了Morozov偏差方程,将图像退化模型的Tikhonov正则化解构造Morozov偏差方程.用小波变换估计图像信号的噪声能量,并运用到Morozov偏差方程中,利用Morozov偏差原理迭代求取正则参数,从而得到正则化解.提出了一种具有噪声能量估计的偏差迭代图像复原正则化方法.实验表明:在图像信号噪声未知的情况下,该迭代方法的图像复原具有较快的计算收敛速度和较好的复原能力.

Restoration of discrepancy iterative images using mechanism for evaluating noise

Morozov discrepancy equation was investigated and formulated by Tikhonov regularization solution that was derived from the degraded image model.The image noise,as the knowledge a prior,was estimated by employing wavelet transform and the equation was modified.A novel method that seeks the optimal regularization parameter by iteration with the improved Morozov discrepancy equation was proposed.The experimental results show that under a lack of the noise information of the observed image the iterative convergence speed and restoration effect of the proposed method were obviously superior to the Van Cittert iterative algorithm.The iterative speed of the proposed algorithm was thrice to quintuple as fast as the Van iteration and PSNR(peak signal-noise rate) was 1 to 5 dB bigger than Van method.As a conclusion the proposed image restoration method is applicable and promising.