Structural vibrations of bearing housings are used for diagnosing fault conditions in bearings, primarily by searching for characteristic fault frequencies in the envelope power spectrum of the vibration signal. The fault frequencies depend on the non-stationary angular speed of the rotating shaft. This paper explores an imaging-based approach to achieve rotational speed independence. Cycle length segments of the rectified vibration signal are stacked to construct grayscale images which exhibit unique textures for each fault. These textures show insignificant variation with the rotational speed, which is confirmed by the classification results using their local binary pattern histograms.