For the problem that the gear fault feature is difficult to extract in actual working condition,a fault feature extraction method based on vibration signal time-frequency image is proposed,which is based on the global texture of the contourlet transform and the local texture of the local binary pattern. Firstly,the vibration signal is transformed into the time-frequency domain by wavelet transform and then the time-frequency gray image is obtained. Then,contourlet transform is performed on the gray image,the low frequency and high frequency subbands are gotten,the low-frequency mean,standard deviation and the high frequency subbands of each layer of the average energy are extracted as part of a feature vector. At the same time,the feature values of the local binary pattern of the time-frequency gray image are extracted and another feature vector is obtained,and the final feature vectors are obtained by the combination of the two feature vectors. Finally,the data of different conditions from a gearbox and rolling bearing are classified and tested by using SVM,the experimental results show the effectiveness of the method,an effective method for the pattern recognition of mechanical equipment is provided.