A new method is proposed to improve the accuracy of method for classifying transient power quality disturbance. First, the grayscale images are created by normalizing the data of disturbance voltage waveforms. Then image enhancement techniques, such as gamma correction and edge detection as well as peak detection methods, are employed to produce new grayscale images and binary images so as to make characteristics of the disturbance striking. The morphologic feature values are extracted from the binary images. At last, the probability neural network (PNN) is trained by the morphologic feature values and then used to classify transient power quality disturbance. The proposed new method for classifying transient power quality disturbance is simulated based on numerical examples. Simulation results show that the accuracy of the new method is better than the existing methods, it is effective and practical.