A method enabling a destroyed part of a joint or joint cartilage to be extracted accurately and at a high speed with a good repeatability and enabling quantitative and simple 3D analysis of the joint and the destroyed part, that is, a method of 3D image processing comprising filling in a medullary cavity region, comprised of a hollow region inside a joint, of a digitalized image of a cross-section of the examined joint using the Expansion and Shrinkage method, performing 3D labeling by a 3D image obtained by stacking digitalized images of cross-sections of the examined joint generated at a step of extracting a contour of the cross-sectional image of the joint or digitalized images of cross-sections of the examined joint not pre-processed, and defining the joint image to be evaluated.