An image processing device is provided that can precisely measure the photographing position or posture of a photographing device or the coordinates of an object based on sequentially changing photographed images. A series of sequentially photographed images are acquired, from which feature points are extracted. The feature points are tracked and correlated to each other. Stereo images are selected from the series of photographed images correlated, and subjected to an orientation and a 3D measurement. The error range of corresponding points obtained by the 3D measurement is calculated. Based on the calculated error range, it is determined whether or not the corresponding points are appropriate for 3D measurement. Those corresponding points determined as inappropriate are deleted. Another orientation and another 3D measurement are performed using those corresponding points excluding the deleted ones, thus improving the measurement precision.