The computer calculates the relationship between the ear shape data indicating the difference between the point cloud representing the ear three-dimensional shape and the point cloud representing the reference ear three-dimensional shape, and the principal component weight vector indicating the weight of each principal component of the ear shape data. Ear shape data is generated by assigning a principal component weight vector to the ear shape model shown, and an estimated 3D shape of the target ear corresponding to the target ear image represented by the image data is specified from the generated ear shape data.コンピュータが、耳の立体形状を表す点群と基準耳の立体形状を表す点群との差分を示す耳形状データと、耳形状データの各主成分の重みを示す主成分重みベクトルとの関係を示す耳形状モデルに、主成分重みベクトルを付与することで耳形状データを生成し、画像データが表す目標耳画像に対応する目標耳の推定立体形状を、生成した耳形状データから特定する。
