Let's suppose that there were only 1,000 ravens in the whole world, and you saw 999 of them chosen at random, and those were black. That would be very STRONG evidence that all ravens are black.
Let's suppose there were 1,000,000 ravens, and you saw 999 chosen at random, and all were black. Well, that'd still be pretty good evidence to intuit that blackness is a general property of ravens, though the existence of albino members of other species should probably make you qualify that 'all.'
Now, how many "non-black" objects are there? Waaaaaaaaay more. You COULD prove that all ravens are black by checking every single non-black object in the universe. But a random sample of non-raven objects is so non-representative of that huge set that it is ridiculously weak evidence. Induction works best when you have entities that could reasonably be expected to be similar to each other on some level. It's easier to make generalizations about "dogs" than it is to generalize about "things represented by words that begin with d" or "things that are not cats."
This doesn't mean that "all ravens are black" is 'easier to prove' than "all non-black things are not-ravens". The two statements really are logically equivalent. But it's more efficient to look at the ravens than it is to sample the the rest of the universe.