Vibrothermography (Sonic IR, thermosonics) is a technique for finding cracks through frictional heat given off in response to vibration. Vibrothermography provides a sequence of infrared images as output of the inspection process. A fast and accurate automatic crack‐detection algorithm for the sequence‐of‐images data will greatly increase the productivity of vibrothermography method. Matched filtering is a technique widely used in signal detection, and it is the optimal linear filter to maximize the signal‐to‐noise ratio in the presence of additive uncorrelated stochastic noise. Based on key features from images of known cracks, we can construct a three‐dimensional matched filter to detect cracks from the vibrothermography data. In this paper, we evaluate the matched filter developed from a vibrothermography inspection sequence‐of‐images. The probability of detection for the matched filter detection algorithm is then compared with the probability of detection for a simpler detection algorithm that is based on a scalar measure of the amount of heat generated in an inspection. Our results show the matched filter algorithm provides improved detection capability when a flaw signature is known approximately.