Improved mesh models of images through the explicit representation of discontinuities

A new mesh model for images that explicitly represents image discontinuities (i.e., image edges), called ERD, is introduced. Then, two mesh-generation methods, known as ERDED and ERDGPI, that select the parameters of this new model for a given image are proposed. The proposed mesh-generation methods are shown to be capable of producing image approximations of higher quality than other competing strategies, both in terms of squared error and subjective quality.