Dual Tree Complex Wavelet based Regularized Deconvolution for Medical Images

Deblurring in the presence of noise is a hard problem, especially in Ultrasound and CT images. In this paper, we propose a hybrid approach of wavelet-based image deconvolution that incorporates Fourier-domain system inversion followed by wavelet-domain noise suppression. In contrast to conventional wavelet-based deconvolution methods, the algorithm employs a regularized inverse filter to operate when the system is non-invertible as well as exploits the properties of dual tree complex wavelet transform (DT-CWT) to remove blur and noise without the need for assuming a specific noise model. Furthermore, the proposed approach uses an adaptive shrinkage function based on median, mean and standard deviation of wavelet coefficients to suppress noise while preserving the sharpness of the image. Its application on ultrasound and CT images has shown a clear improvement over alternative methods.