Classification images (CIs) can reveal observers' strategies in a variety of visual tasks. However, one weakness of the CI method is that many trials are needed to obtain stable data (Sekuer etal., 2004). We examined whether CIs can be obtained with fewer trials, thereby making it possible to use the method with clinical populations. With not entire but sampled faces presentations, we obtained CIs from seven typical and eight autistic observers. Although normal, or raw, CIs based on 1,450 trials could not show face processing strategy of individual observers, the squared CIs after smooth filtering did clearly show it. Typical observers strongly used the region of eyes and eyebrows. In contrast, autistic observers did not strongly used that region, but some of them used forehead area.