K-mer counting of biological sequence is a fundamental and very important problem in biological information processing. This paper focuses on counting k-mers at each position of multiple sequences within aligned mode. We present a new backward traverse k-mer counting algorithm called BTKC. BTKC algorithm takes full advantage of the k+1-mer's statistic information to obtain k-mer's statistic information quickly. Thus, it's no need to traverse the whole sequences when counting each single k-mer. Both the algorithm's time complexity and experiment results show that BTKC gets an obvious improvement compared with forward traverse k-mer counting algorithm FTKC, and its time complexity was found not to be realted with the range of k-mer length.