Image Resolution Enhancement in Wavelet Domain

Resolution enhancement of pictorial data is desirable in many applications when images at desirable resolution levels are not available. It is a classic signal interpolation problem and several conventional approaches such as zero-order interpolation, bilinear and spline interpolation are widely used. Recently several efforts in the field have utilised wavelet-domain methodologies with the intention of overcoming some of the problems associated with conventional treatment. In this work, three wavelet domain image resolution enhancement algorithms are proposed. The first proposed algorithm is based on the estimation of detail wavelet coefficients at high resolution scales. The second algorithm utilises the cycle spinning methodology. For the third algorithm a directional variant of the cycle spinning methodology is proposed with an aim of reducing the over-smoothing of salient image features as well as offering a reduction in computational complexity. The results show that the proposed methods are considerably superior to conventional image interpolation techniques, both in objective and subjective terms, while also comparing favourably with the competing methods.