Progressive Image Reconstruction based on Multi-scale Edge Model

In this paper, we present a progressive image reconstruction scheme based on the semantically scalable multi-scale edge representation of images, with the resolution and visual quality scalable to various bitrate requirements. In the multi-scale edge representation an image is decomposed into its multi-scale primal sketch and the background where the multi-scale primal sketch preserves the structural semantics of images, and the background represents the smooth locale. Edge compensation is performed to smoothly remove edges at each scale. The multi-scale edges are then embedded encoded using the GFA modeling. The image reconstruction is progressively achieved by synthesizing multi-scale edges on the reconstructed image obtained from previous scale. As edge synthesis is performed at consecutive scales, the visual quality of the reconstructed image is progressively enhanced. Experiment shows that the proposed scheme performs well at low bit-rate multiresolution representation and progressive reconstruction