An image processing apparatus comprises: a first feature quantity calculating unit that calculates, for each of the pixels of an image obtained by image-capturing a living body tissue, a first feature quantity indicating an index as to what a local area, which includes the pixel of interest and pixels in the vicinity of the pixel of interest, is shaped like a second feature quantity calculating unit that calculates, for each of the pixels of the image, a second feature quantity indicating a distribution state of gradient directions in the local area an evaluation value calculating unit that calculates, for each of the pixels, on the basis of calculation results of the first and second feature quantities, a shape evaluation value that is used to allow a predetermined shape structure included in the image to be determined and an area extracting unit that extracts, on the basis of a calculation result of the shape evaluation value, from the image, a candidate area in which the predetermined shape structure is presumed to exist.