A digital still camera includes a CCD image sensor for photographing an object image by photoelectric conversion, to obtain image data. A face detector determines a face feature value of a human face at an object image by image recognition according to the image data, and detects the face. A stability checker monitors the face feature value, and outputs stable result information when the face feature value is within a prescribed range consecutively for time of a predetermined length or consecutively for a predetermined number of times. A controller automatically starts image pickup of the CCD image sensor when the stable result information is output by the first stability checker.