Image restoration using SVD and adaptive regularization

Abstract EN

In this paper, we consider the problem of image restoration degraded by a shift invariant blur and corrupted by random additive noise.

A new approach is implemented by using the Singular Value Decomposition (SVD).

In absence of noise, the image restored is the same as the original.

If, in the other hand an additive noise is added to the degradation model, we propose an iterative algorithm which attempts to maximize the SNR of the restored image.

In all experiments, colour images are considered in RGB distribution, where each subspace can be regarded as a grey image space and is processed separately.

The quality of the obtained results attests the efficiency of the proposed method.