The purpose of this study was to develop a computerized method for estimation of the location of a lung tumor in cine images on an electronic portal imaging device (EPID) without implanted markers during stereotactic body radiotherapy (SBRT). Each tumor region was segmented in the first EPID cine image, i.e., reference portal image, based on a multiple-gray level thresholding technique and a region growing technique, and then the image including the tumor region was cropped as a 'tumor template' image. The tumor location was determined as the position in which the tumor template image took the maximum cross-correlation value within each consecutive portal image, which was acquired in cine mode on the EPID in treatment. EPID images with 512 x 384 pixels (pixel size: 0.56 mm) were acquired at a sampling rate of 0.5 frame s(-1) by using energies of 4, 6 or 10 MV on linear accelerators. We applied our proposed method to EPID cine images (226 frames) of 12 clinical cases (ages: 51-83, mean: 72) with a non-small cell lung cancer. As a result, the average location error between tumor points obtained by our method and the manual method was 1.47 +/- 0.60 mm. This preliminary study suggests that our method based on the tumor template matching technique might be feasible for tracking the location of a lung tumor without implanted markers in SBRT.