SUPER-RESOLUTION USING EXAMPLE-BASED NEURAL NETWORKS

摘要

PURPOSE: A method for enlarging high definition image using neural networks model based on an example is provided to approximate information about high frequency component through back propagation neural network. CONSTITUTION: Division to and image of a low frequency component is performed. Division to an image of mid frequency component which very low frequency is removed is performed. Division to an image of high frequency is performed. The image of the high frequency component is obtained through difference of each pixel value with the image of low frequency component and the original image of high definition.