In this paper, we propose a novel method termed as the histogram and ultrafast segmented model identification of linearity errors (H-USMILE), in order to implement an efficient linearity test of ADCs. We use the standard histogram method with only a small number of test samples to obtain coarse INL results, along with the segmented non-parametric model of the ultrafast segmented model identification of linearity errors (USMILE) algorithm employed to convert the coarse INL results into precise ones. Experimental results show that, by using the proposed method, we can significantly reduce the required test data and test time while maintaining or even improving the precision of the standard histogram method.