Gravar-mail: Feature-Based Attention in Early Vision for the Modulation of Figure–Ground Segregation