Time series motifs are frequently occu"ing but previously unknown subsequences of a longer time series. This motif concept is generalized to k-motifs problem, where the top K-Motifs are returned. Discovering time series motifs is a crucial task in time series data mining. Among a dozen algorithms that have been proposed for discovering time series motifs, the most popular algorithm is random projection. However, it still has some drawbacks. In this paper, the authors propose a novel approach for discovering K-Motifs in a long time series with the support of a multidimensional index structure, R*-tree and the idea of early abandoning. The method is disk efficient because it only requires a single scan over the entire time series. the authors demonstrate the effectiveness of the approach by experimenting on real datasets from different areas. The experimental results showed that the proposed algorithm outperforms the most popular method, random projection, in efficiency.