In the image processing apparatus, the control unit generates a plurality of divided regions by dividing the image in accordance with similarity calculated based on pixel values and pixel locations, and identifies a selection range constituted by one or more of the plurality of divided regions in accordance with the selection operation received by the input unit. The control unit performs erosion processing with respect to the selection range by reducing a number of pixels constituting the selection range, and performs dilation processing with respect to the selection range resulting from the erosion processing by increasing the number of pixels constituting the selection range resulting from the erosion processing. The number of pixels constituting the selection range resulting from the dilation processing is greater than the number of pixels constituting the selection range before the erosion processing.