A subject tracking apparatus which performs subject tracking based on the degree of correlation between a reference image and an input image is disclosed. The degree of correlation between each of a plurality of reference images based on images input at different times, and the input image is obtained. If the maximum degree of correlation between a reference image based on a first input image among the plurality of reference images and the input image is equal to or higher than a threshold, a region with a maximum degree of correlation with a first reference image is determined as a subject region. Otherwise, a region with a maximum degree of correlation with a reference image based on an image input later than the first input image is determined as a subject region.