Roller bearing is one of the most widely used and easily damaged elements in rotary. Fractal dimensions could describe the complexity and irregular in bearing vibrating signal. Fractal dimension based on mathematical morphology has advantages in calculating speed and accuracy, which could distinguish roller bearing status and estimate its faulted behavior accurately. The calculating method of mathematical morphological fractal dimension was stated in the paper, In calculating procedure of fractal dimension, it always lacks guidance for choosing the length of flat structuring element. In allusion to this issue, a new calculating method named sectionalized fractal dimension was proposed in this paper. Applying this method into evaluating and analysis of vibration signal for different faulted roller bearing, it is obvious that the scientificity and preciseness for fractal dimension is improved and it is effective in roller bearing fault pattern recognition.