what the mean of image_size, label_size and stride about srcnn-tensorflow

[mohamedhaiham94]

according to the paper to generate sub images they set image width to 33x33 and stride to 14 , and i think that label size is depended on the padding type so if the padding = "Valid" after first layer the image size will be 25 x 25 x 64 according to this equation ( (n + 2p - f / s) + 1 ) where n is image size , p = padding, f = filter size and s = stride ==> 33 + (2 * 0) - 9 / 1 + 1 = 25 after the second layer applying same equation image size = 25 x 25 x 32 at the last layer image size = 21 x 21 x 1 so according to that label size = 21 but if the padding = "same" i think label size will be 33 instead of 21