Conventional radar imaging systems use Fourier transform for image formation, but due to the target's complicated motion the Doppler spectrum is time-varying and thus the reconstructed image becomes blurred even after applying standard motion compensation algorithms. Therefore, sophisticated algorithms such as polar reformatting are usually employed to produce clear images. Alternatively, Joint Time-Frequency (JTF) analysis can be used for image formation which produces clear image without using polar reformatting algorithm. In this paper, a new JTF-based method is proposed for image formation in inverse synthetic aperture radars (ISAR). This method uses minimum entropy criterion for optimum parameter adjustment of JTF algorithms. Short Time Fourier Transform (STFT) and Fractional Fourier Transform (FrFT) are applied as JTF for time-varying Doppler spectrum analysis. Both the width of Gaussian window of STFT and the order of FrFT, α, are adjusted using minimum entropy as local and total measures. Furthermore, a new statistical parameter, called normalized correlation, is defined for comparison of images reconstructed by different methods. Simulation results show that α-order FrFT with local adjustment has much better performance than the other methods in this category even in low SNR.