An image processing apparatus comprises a first calculation unit for calculating a shake correction amount for a moving image by using a shake signal output from shake detection unit, a second calculation unit for calculating a shake correction amount for a still image by using a shake signal output from the shake detection unit, a generation unit for generating, based on a comparison result between the shake correction amount for the moving image and the shake correction amount for the still image, evaluation values serving as indices used for extracting a still image from among frame images of the moving image, and a recording unit for recording the evaluation values in association with the frame images of the moving image.