Image upscaling system, training method thereof, and image upscaling method

An image upscaling system comprising:

• a first convolutional neural network module;
• a second convolutional neural network module;
• and
• at least one synthesizer;
• wherein the first convolutional neural network module and the synthesizer are alternately connected to one another and the synthesizer is connected between the first convolutional neural network module and the second convolutional neural network module;
• the first convolutional neural network module is configured to:
• receive an input image and a first supplemental image having a same resolution as the input image, to generate a first number of feature images based on the input image and the first supplemental image, and to output the first number of feature images to the synthesizer connected thereto;
• the second convolutional neural network module is configured to:
• receive an output image from the synthesizer and a second supplemental image having a the same resolution as the received output image, to generate a second number of feature images based on the output image from the synthesizer and the second supplemental image, and is operable to output the second number of feature images to one of:
• a next synthesizer connected thereto or an output of the image upscaling system;
• wherein the synthesizer is configured to synthesize every n*n feature images in the received feature images into one feature image, and to output a third number of feature images synthesized to one of:
• the second convolutional neural network module connected thereto or as an output of the image upscaling system;
• and
• wherein n denotes an upscaling magnification factor of the synthesizer, which is an integer greater than 1, and the number of feature images received by the synthesizer is a multiple of n*n.