enlarge the pictures with an equal ratio vertically and horizontally. Also, enlarge the dimensions by a multiple of the current dimension.
a 100x100 image may be resized to 200x200 perfectly. You may also reduce the image back from 200x200 to 100x100 and get the original image. This is not guaranteed if you do something like enlarge a 100x100 to 150x150 then go back again. Other multiples can be increasing it to 300x300, 400x400 etc...
This way the program doesn't have to do various visual cheats to make hte image look good when enlarging it. If you take a 100x100 image and increase the size to 200x200, then all the program has to do is change each pixel to 4 pixels. When reducing the image back again half it's current size, it'll take each 4 pixel square and reduce it to a single pixel. No pixels are lost, guaranteed.
another example:
123x456 => 246x912 => 369x1367