High resolution (HR) images have more information about images than low resolution (LR) images. HR image can be more easily recovered from multiple low resolution images than the single LR image. Several LR images hold diverse details and can be dubbed together to get a HR image. Reconstruction of HR image using single LR image with fewer details remain more challenging. This paper proposes a neural network based model to increase the resolution based on feature extraction from a single image. Each LR and HR image is divided into smaller blocks and the histogram as a feature is calculated for each block. Back propagation neural network [BPNN] is used for learning of histogram features of LR image. The simulation has been performed on a dataset of brain MRI images and results show that the obtained neural network model has improved the PSNR and RMSE.
Software And Hardware