I think you can use eXpress (http://bio.math.berkeley.edu/eXpress/overview.html) to do so. If you transcripts from transcritome assembly then map the reads back to transcripts and eXpress and generate the count based statistics.