A de-warp map is generated by applying principal component analysis (PCA) to vectors describing aspects of identified features of an object in an image. PCA provides vectors and coefficients describing curvature or image warping at selected points in the image. Estimates of the warping of the image generally are generated by interpolation and/or extrapolation from the vectors and coefficients provided by PCA. In some applications only two features need be identified. For example, the complicated curvature of the facing pages of an open book can be characterized by two vectors describing positions of top and bottom edges of the book. In such applications PCA can reduce to vector subtraction to determine a basis vector, vector addition and scaling to determine an average vector and simple assignment of known coefficient values. The de-warping map can be used to generate a de-warped version of the image.