The bispectrum is crucial for description of non-Gausssian and/or non-linear signals. In this paper we propose wavelet-thresholding estimators of the bispectrum of zero-mean, non-Gaussian, stationary signals. It is known in the case of Gaussian regression that wavelet estimators outperform traditional linear methods if the regularity of the function to be estimated varies substantially over its domain of definition. The goal of this paper is to extend the wavelet-thresholding estimation method to bispectrum estimation. We will show that, in the context of the bispectrum estimation, wavelet-thresholding estimators outperform linear (kernel) estimators.