If you solved Easy Stego the intended way, then you should be expecting something to do with comparing the individual pixels with their counterparts.

That is good.

Now, Hard Stego is a little bit more involved that Easy Stego. In Hard Stego, every pixel is shifted by some small increment of RGB, like [0,4,0], [0,6,8], or [1,1,5].

Let's say you've gotten this far, and have gotten all off these change arrays. Now, you have to figure out how I used this information to encode data into the image.

The correct thing to think of would be ASCII values.

See, if you convert the RGB arrays straight into integers (like [1,2,1] -> 121), then you'll get a lot of integers less than 128. Personally, for CTFs, I think that it's never a bad idea to try converting smaller integers into characters via ASCII - it takes about 2 seconds, so even if it's useless you haven't wasted much time.

Some of your numbers will be negative - if you're stuck on that, put yourself in my shoes. Let's the Doge's eyes have a pixel with the value [255,255,255]. Now, I have to encode the "space" character (#40) into that pixel. What do I do? I can't make the Green of the pixel go to 259 - instead, I'll just have it wrap around. Now, in the edited picture, you'll be seeing a pixel with RGB value [255,4,255], so when you subtract you'll get [0,-251,0]. Therefore, you should just add 256 to any negative integers you get.