We present a novel and effective approach to segmenting overlapping cytoplasm of cells in cervical smear images. Instead of simply combining individual cytoplasm shape information with the intensity or color information for the segmentation, our approach aims at simultaneously matching an accurate shape template for each cytoplasm in a whole clump. There are two main technical contributions. First, we present a novel shape similarity measure that supports shape template matching without clump splitting, allowing us to leverage more shape information, not only from the cytoplasm itself but also from the whole clump. Second, we propose an effective objective function for joint shape template matching based on our shape similarity measure; unlike individual matching, our method is able to exploit more shape constraints. We extensively evaluate our method on two typical cervical smear data sets. Experimental results show that our method outperforms the state-of-the-art methods in term of segmentation accuracy.