Wavelet domain image super-resolution reconstruction based on image pyramid and cycle-spinning

On the base of the analysis of single image super-resolution reconstruction, wavelet domain super-resolution reconstruction based on image Pyramid and cycle-spinning is proposed in the paper. The algorithm consists of the following parts: the predictive high-resolution image are obtained by a LR image based on laplacian pyramid method; the discrete wavelet transform is used for predictive high-resolution image to obtain the high-frequency wavelet coefficients; a high-resolution image is obtained by the inverse wavelet transform; Finally the cycle-spinning method is used to reduce ringing. The proposed algorithm avoids the application of iterative method, reduces the complexity of calculation and applies to large dimension low-resolution images. The results show the proposed approach has succeeded in obtaining a high-resolution image with a high peak signal-to-noise ratio and good visual quality.