By using various information such as GPS information, cartographic information, and atmospheric pressure acquired by a sensor, a CPU of an imaging device calculates a motion vector in a limited MV calculation area acquired by excluding from images an area where radial motion vectors occur which highly possibly give an adverse effect to motion vector calculation, in accordance with the gravity direction with respect to the imaging device, the orientation of the imaging device, and the slope angle of the road where the imaging device is moving forward. After calculating the motion vector, the CPU estimates a blur amount between frames from the calculated motion vector and, based on this blur amount, performs the alignment of a feature point between the frames by image deformation processing.