The authors have developed a target speed estimation method for an ultra-wideband radar imaging system. The method is able to generate a focused image even if the target is mov- ing at an unknown speed during the measurement, where many other techniques produce blurred images. Our technique uses the cross-range blurriness as a metric to measure the image focusing, and select the best-focused image among many generated for various candidate speeds. The target speed is estimated by finding the minimum cross-range blurriness, and the method com- pensates for the estimated speed to generate a focused three-dimensional image automatically. By exploiting the fact that the head size differs relatively little among individuals, we detected the head position in a radar image, and calculated the blurriness around the region to evaluate the focuse of the image. The speed estimation method is applied to the measurement of a man- nequin on a moving stage, and it is shown that the target speed is estimated with an accuracy of 5%. Application results show that the developed technique can estimate target speeds accurately, and produce an image with sufficient resolution for the intended application.