According to one embodiment, an image processing apparatus includes a first setting unit, a second setting unit, and a specifying unit. The first setting unit detects a position of at least a part of an object in an image so as to obtain, for one pixel or each of a plurality of pixels in the image, a first likelihood that indicates whether the corresponding pixel is included in a region where the object is present. The second setting unit obtains, for one pixel or each of a plurality of pixels in the image, a second likelihood indicating whether the pixel is a pixel corresponding to a 3D body by using a feature amount of the pixel. The a specifying unit specifies a region, in the image, where the object is present by using the first likelihood and the second likelihood.