Abstract
With an ever-increasing demand for very low bit rate image compression scheme, classical image coding methods start to show their inability to represent images with acceptable quality at this bit rate. A distinct image compression scheme is being explored that does not rely on the usual additive decomposition of a source image into set of predefined spanning functions. This proposed scheme only encodes implicit properties of the image and reconstructs an estimate by constrained highpass energy minimization. This has particular advantages for very low bit rate perceptually oriented image compression. We show experimental results in a context where we mainly target the smooth areas and the edges of the image.