Abstract: Entropy-based divergence measures have provided an impelling tool in evaluating sequence complexity, predicting CpG island, and detecting borders between coding and non-coding DNA regions etc. In this paper, two new divergence measures: the α-KL divergence and the α-Jensen-Shannon divergence were defined and a coarse-graining vector of amino acids- corresponding codons was proposed according to codons GC-content, in order to improve the computational approach to finding borders between coding and non-coding in rice. By comparing the accuracies gained by different vectors (the Jensen-Shannon divergence, the Jensen-Renyi divergence, the α-KL divergence and the α-Jensen -Shannon divergence) for detecting borders between coding and non-coding DNA regions, it was showed that recognition efficiency of the new information measures with the coarse-graining vector was increased by 4~5 times for the ‘stop codon’ of coding regions in rice, when compared with Bernaola’s method.