Abstract
Regularized restoration is one of the powerful image restoration techniques because it preserves image details with a high degree of fidelity in the restored image. The main problem encountered in regularized image restoration is the evaluation of the regularization parameter. There are several methods for the evaluation of this parameter which require knowledge of the noise variance in the degraded image. After evaluating this parameter, regularized restoration is implemented by applying a regularization filter on the degraded image. In this paper, we propose a new iterative method for the evaluation of this parameter. This method depends on the maximization of the power in the restored image by the coincidence of the passband of the regularization filter with the frequency band in which, most of the image power exists. The suggested method doesn’t require a priori knowledge of the noise variance. Results show that the estimated value of the regularization parameter leads to a minimum mean square restoration error.
Highlights
Digital image restoration is a problem which has been extensively treated in the literature [1-12]
The suggested method for evaluating the regularization parameter is tested for two different images with different spatial activities; the Cameraman and the Mandrill images illustrated in Figs. (8) and (9), respectively
For the Mandrill image which is of a high frequency nature, the MEAN SQUARE ERROR (MSE) reaches its minimum value with the maximum value of the total power in the restored image and keeps this minimum value for a long range of λ
Summary
Digital image restoration is a problem which has been extensively treated in the literature [1-12]. The main objective of image restoration is to obtain a good estimate of the original image from a degraded image. Degradations in images have several origins such as out of focus blurring, linear motion blurring and Gaussian blurring. These types of blurring can be modeled as lowpass filters affecting the original image [13]. The image restoration problem is a deconvolution problem. The existence of noise in the degraded image increases the difficulty of the image restoration process. The problem of deconvolution in the presence of noise is classified as an ill-posed inverse problem [6]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have