EMM盲运动模糊图像的恢复

盲图像恢复的主要困难是信息不足,而为了恢复图像和确定点扩散函数需要适当的先验知识。解决这个问题的方法有ML法、EM法以及正则化方法等。但是这些方法的计算量都太大,针对上述方法的不足,文章提出了一种恢复图像的新算法,它通过恢复残差的最小化和后验概率的最大化来估计参数和恢复图像。其中,巧妙地利用了最陡梯度法和共轭梯度法的迭代求解。对由于运动造成的模糊图像,可以明显地改善图像的质量,实验结果证明,在对模糊操作没有严格限制的情况下,仍可得到较好的恢复图像。

EMM Motion Blured Image Restoration

The primary difficulty with blind ima ge restoration is insufficient information.This calls for proper incorpora-tion of a priori knowledge about the image and the point-spread function(PSF).w ell-know space-adaptive regularization method and EM method for image restorat ion are extended to address this problem.A new method is proposed to restore bl ur image using ML and Least Squares method.It use minimize a cost function an d maximize PSF function to restore the image and identify the parameters.This m ethod use steepest descent and conjugate gradient methods to minimize the cost function and maximize posteriori probability to improve the blur image,Which is also to improve the image quali-ty in some degree with no strict limitation .