Segmentation is an important preprocessing step for pattern classification in financial time series. In this paper, we propose a novel segmentation method called Perceptually Important Point with Binary tree (PIP-Btree) for efficient preprocessing of financial time series for classifying chart patterns. PIP-Btree takes advantage of a standard binary tree for improving the efficiency without compromising the accuracy of original Perceptually Important Point (PIP) method. Besides, attribute parameters of PIP-Btree support self-updating when a new data point arrives in a streaming time series. In the experiments, efficiency of PIP-Btree is compared to other segmentation methods. Accuracy of PIP-Btree method is also evaluated for financial chart pattern classification after it is integrated with a rule-based pattern matching approach. Experiment results reveal that PIP-Btree achieves best score in both efficiency and accuracy performance.