An image processing apparatus according to the present disclosure includes an illuminating section which sequentially irradiates an object with first and second illuminating light beams polarized in first and second directions. First and second polarization images are generated based on signals representing light transmitted through polarizers having the polarization transmission axis in respective directions that are parallel to, and intersect with, the first direction while the object is being irradiated with the first illuminating light beam, and third and fourth polarization images are generated based on signals representing light transmitted through polarizers having the polarization transmission axis in respective directions that are parallel to, and intersect with, the second direction while the object is being irradiated with the second illuminating light beam. A depressed object surface region is detected based on the first and second polarization images and/or the third and fourth polarization images.