This work proposes an adaptation of the Empirical Mode Decomposition Time-Frequency Distribution (EMD-TFD) to non-analytic complex-valued signals. Then, the modified version of EMD-TFD is used in the formation of Inverse Synthetic Aperture Radar (ISAR) image. This new method, referred to as NSBEMD-TFD, is obtained by extending the Non uniformly Sampled Bivariate Empirical Mode Decomposition (NSBEMD) to design a filter in the ambiguity domain and to clean the Time-Frequency Distribution (TFD) of signal. The effectiveness of the proposed scheme of ISAR formation is illustrated on synthetic and real signals. The results of our proposed methods are compared to other Time-Frequency Representation (TFR) such as Spectrogram, Wigner-Ville Distribution (WVD), Smoothed Pseudo Wigner-Ville Distribution (SPWVD) or others methods based on EMD.