Image Feature Enhancement Using Hybrid Superpixel Reconstruction

Feature enhancement is a most fundamental and challenging issue in machine vision. In this article, a feature enhancement method is presented using hybrid superpixel reconstruction. (1) A superpixel concept, called hybrid superpixel, is proposed for over-segmentation. Second, fast feature enhancement (FFE) is developed based on hybrid superpixel, fuzzy optimization and data reconstruction. Extensive experiments are performed, showing that FFE advances the state-of-the-art comparison systems significantly.