This paper proposes a partial discharge (PD) pulse image representation method, which converts the sampled time-domain signal into image, intuitively representing the processing process, and facilitating further partial discharge pulse signal analysis, for example, feature extraction and recognition. After study of the image characteristics of partial discharge electromagnetic pulses, time-frequency transforms are used as a method of image representation, and three types of algorithms, short-time Fourier transforms, S-transforms, and Wigner-Ville distribution are compared for partial discharge pulse and background signal feature representation in factors of the sharpness of the contour of the pulse waveform, the energy difference between the partial discharge pulse and the background signal, and the shape contrast effect. The application based on simulation data and model test data shows that Compared with direct time-domain analysis, the PWVD algorithm can effectively represent pulse characteristics in strong noise environments.