Hybrid image restoration using SWT based denoising and regularization in frequency domain

摘要

This paper presents a hybrid approach for image restoration which is degraded from known estimated blurring kernel, read out noise and photon counting noise. The novelty of the approach lies that it does not require noise power at the time of deconvolution while it accounts all possible noise, which occurred at the time of present acquisition of image. In the model, first stage is denoise the degraded image with use of Stationary Wavelet Transform (SWT) using soft threshold. And in second stage image regularization by maximum aposteriori. Test is performed on various real life single frame degraded image with noise. The approach shows the superiority of the scheme in term of subjective quality and increase in signal to noise ratio.