Famous feature point descriptors such as SIFT and SURF allow reliable real-time matching but at a computational cost that limits the number of points that can be handled on PCs, and even more on less powerful mobile devices. A recently proposed technique called Binary Robust Independent Elementary Features (BRIEF) uses binary string as an efficient feature point descriptor. It has been shown that it is highly discriminative even when using relatively few bits. However, we note that the resulting features are not robust when the intensity difference between two compared pixels is small according to the test definition of BRIEF. In this paper, we try to remedy this and replace the test definition with a slightly modified one. Our experimental results show that the modified descriptors are more distinctive and more robust to typical image disturbances such as viewpoint change and image blur that occur in real-world scenarios. We also highlight its effectiveness by incorporating it into a SLAM system called Parallel Tracking and Mapping (PTAM) and demonstrating substantial performance increases.