Restoration of color images using wavelets

Inverse filtering and Wiener filtering are two classical approaches used for image restoration. The inverse filtering approach is very sensitive to image noise due to its high-pass nature. Wiener filtering can be interpreted as an inverse filtering step followed by a noise attenuation step. One of the issues is that it requires a priori estimation of the power spectrum of the noise in the corrupted image. This paper presents a wavelet-based scheme for the restoration of color images. The scheme consists of two steps: inverse filtering and wavelet based de-noising. The Daubechies wavelet is employed to transform the data into a different basis where a large number of coefficients correspond to the noise whereas the signal is restricted to a few coefficients. The de-noised data is obtained by inverse-transforming the suitably threshold coefficients. A set of experiments was designed where the two above steps were interchanged to study the effect on the restoration process. The test images were created upon corrupting color image data with directional motion blur and additive Gaussian noise. All the algorithms were designed and tested in the MATLAB™ environment. The new approaches were compared with the classical image restoration approaches, on the basis of the mean square error (MSE) criterion. The experimental results showed qualitatively and quantitatively that the wavelet-based technique outperforms the classical methods.