An image encoding method that encodes an image so as to optimize estimated values of processing units which are a coding unit, a prediction unit, and a transform unit by selecting sizes of the individual processing units. The method includes a division step that divides an image having a size of the coding unit into a plurality of blocks; an activity computation step that computes activities pertaining to the individual blocks; and a transform unit size determination step that determines, based on the computed activities, a size pattern of the transform unit within the image having the size of the coding unit.