While that would result in decreased entropy that a decent general-purpose compression algorithm like LZ77 could take advantage of, this preprocessing step is lossy and probably not worth the bytes it would save you.