Singular value decomposition based fusion for super-resolution image reconstruction

Singular value decomposition (SVD) has been successfully used in image processing such as image compression, feature extraction and detection. The paper proposes the use of SVD to enhance super-resolution results. The proposed method converts the registered reference image into the SVD domain and then images´ singular values are fused based on the fusion rule before performing the interpolation. The objective of using SVD is to integrate the important features from low resolution images. Simulation results of applying SVD-fusion prior to interpolation show significant performance improvement when compared to standard interpolation techniques and also with the existing learning-based super-resolution approach.