Would you normalise to the length of the chromosome or to the number of genes per chromosome? Or to the total length of genes on the chromosome?
If the OP compare multiple sample, there is no need to normalize for chromosome length or number of genes. But you need some kind of between sample normalization. The easiest (and probably not the best) way is to normalize to the total number of reads mapped.
PS : samtools idxstats is very efficient indeed, much faster than a samtools view command as Buffo proposed in its comment.
Aligning with bowtie at >97% identity and then count; samtools view -S.
it's RNA-Seq so bowtie is not the best aligner to do that. I personaly use STAR to align RNA-Seq
? star also creates a .sam file index doesn´t it?
bowtie2 is not splice aware.