Denoising of Complex Valued Images by Sparse Representation

The data acquired from many complex valued image systems are always disturbed by noise. The proposed denoising method of complex valued numbers is based on the extended sparse representation. The algorithm extends orthogonal matching pursuit(OMP) to the complex domain to do the coding and learns the adaptive dictionary from the patches extracted from the noisy image. We reach the conclusion that the denoising results are related to the sparsity of the representation. The experimental results on the real data show that the proposed algorithm achieves very good results, especially in the peak signal to noise ratio for the phase of the complex numbers.