There are two ways to look at this. First as an embellishment to ordinary communication methods (which includes email). That would be a more entertainment-focused product, which may be interesting.

The other direction, which I hope is not what you are hinting at, is a universal mapping of words to pictures. From what I know, historically and theoretically, this is impossible. For one, universal language attemps have failed (esperanto being the largest ever and it's a dud). For two, not all word forms have concrete representations (try meta-concepts, the concept that represents, and even "meta-concepts" itself). For three, Sapir-Whorf hypothesis. http://en.wikipedia.org/wiki/Sapir-Whorf_hypothesis

Language is a statistical thing; every new utterance communicated adds more data to the pool. Hence, even while there are dictionaries, the very second a definition is written down, if somebody uses the word, the definition is no longer perfect. In conjunction, there will never be a "best match" picture for a word.

If this is really your idea, consider not doing voting on pictures, but rather making a gigantic, associative map of words across languages. That will have immense educational value.