Generally, the video resolution ranges from 720p (image 1280 x 720) to 1080p (image 1920 x 1080), and upsampling is 1.5 times. However, the pixel shuffle is used for upsampling on a common gan network, which does not support non-integer multiples. Are there other functions for upsampling, and which method has better performance and precision?
The best solution for upscaling images is to use a resize function that leverages exponential functions. Essentially, you fit an exponential curve to the data and then resample from the interpolated points on that curve to reconstruct the original signal at the higher resolution. Is there any reason you would avoid this approach?