A lateral shake prevention when a video picked up by a digital camera is reproduced is realized, and viewing on a large screen television set is made comfortable.At the time of the image pickup recording, face detection position center coordinate data is recorded in an attribute information unit of a stream as ARI data for each VOBU (GOP) in units of 0.5 seconds. At the time of the reproduction, the ARI data of GOP stream is collectively read ahead in units of RUV at the time of the reproduction of the video included in detection data of the face center position coordinate to read a large number of VOBUs in advance, an interpolation of missing times in units of GOP with respect to the face center position coordinate is performed, and a screen shift amount is subjected to a time axis smoothing computation through a time moving average computation in a frame unit.