Super resolution neural network with multiple outputs with different upscaling factors

A system for processing images, the system comprising:

• one or more computer storage devices configured to store:
• a first module implementing a super resolution neural network, the super resolution neural network including feature extraction layers and multiple sets of upscaling layers sharing the feature extraction layers, the multiple sets of upscaling layers configured to upscale an image according to different respective upscaling factors;
• and
• a second module configured to further upscale the image;
• one or more processors configured to execute instructions, the instructions causing the one or more processors to:
• receive an input image;
• upscale the input image according to a final upscaling factor F by:
• providing the input image to the first module, the multiple sets of upscaling layers of the super resolution neural network upscaling the input image according to the different respective upscaling factors to produce respective first module outputs;
• selecting the first module output from the set of upscaling layers with the respective upscaling factor closest to the final upscaling factor F;
• and
• if the respective upscaling factor associated with the selected first module output is not equal to the final upscaling factor F, providing the selected first module output to the second module to further upscale the selected first module output to produce a second module output corresponding to the input image upscaled by the final upscaling factor F;
• and
• output:
• the selected first module output if the respective upscaling factor associated with the selected first module output is equal to the final upscaling factor F, or
• the second module output if the respective upscaling factor associated with the selected first module output is not equal to the final upscaling factor F.

One or more non-transitory computer-readable storage media, having computer-executable instructions stored thereon, wherein when executed by a one or more processors, the computer-executable instructions cause the one or more processors to:

• receive an input image;
• upscale the input image according to a final upscaling factor F by:
• providing the input image to a first module, the first module implementing a super resolution neural network, the super resolution neural network including feature extraction layers and multiple sets of upscaling layers sharing the feature extraction layers, the multiple sets of upscaling layers configured to upscale the input image according to different respective upscaling factors to produce respective first module outputs;
• selecting the first module output from the set of upscaling layers with the respective upscaling factor closest to the final upscaling factor F;
• and
• if the respective upscaling factor associated with the selected first module output is not equal to the final upscaling factor F, providing the selected first module output to a second module configured to further upscale the selected first module output to produce a second module output corresponding to the input image upscaled by the final upscaling factor F;
• and
• output:
• the selected first module output if the respective upscaling factor associated with the selected first module output is equal to the final upscaling factor F, or
• the second module output if the respective upscaling factor associated with the selected first module output is not equal to the final upscaling factor F.