The parameter identification of the point spread function is very important for image restoration. Motion blur and defocus blur are the common blur types which degrade image. For the two blur types, a robust bispectrum-based method is proposed to estimate the blur parameters in noisy images. At first, we introduce the two blur types and traditional computational formulas. And then, the bispectrum of two blur types is deduced. In theory, bispectrum is not affected by noise, and corresponds to similar structure of blur function, so we can calculate the blur parameters easily by using the traditional computational formulas. The experimental results show that our method is effective in a certain range. In other words, when the blur parameters are not too large, our method can complete the parameter identification more accurately. When SNR is 25 dB，the deviation of blur parameters recognized is less than 1 pixel in this range.