In this paper, we present an image restoration algorithm that uses multiple captured degraded, low resolution (LR) and noisy images to reconstruct a high resolution (HR) image. For the reconstruction process, a spectral-based blind image deconvolution/restoration technique is proposed. The presented mathematical analysis for the technique is carried out in the frequency domain. The developed analysis is then used to estimate captured images' spectra that are needed for the blind restoration algorithm. Unlike conventional image restoration methods, our spectral-based algorithm: (i) significantly minimizes the effects introduced by additive noise (ii) does not use inverse filtering, which can be unstable, (iii) is efficient in computation complexity when compared to previously reported methods. The proposed algorithm is tested on multiple blurred, LR and noisy medical images. Results show that the proposed algorithm is capable of restoring HR images from degraded observations even at low signal-to-noise energy ratios (SNERs).