Structure Importance-Aware Hidden Images

Hidden images are fascinating as they present viewers a visual phenomenon of perceiving latent illusion from a single image. Recently, several solutions for automatically generating hidden images have been developed in the computer graphics community. While these solutions produce visually appealing hidden results, structure importance of the hidden object is rarely taken into account. In this paper, an automatic algorithm is presented to generate such images by integrating structure importance analysis of the hidden objects. Our algorithm is motivated by the law of closure of Gestalt psychology theory, indicating that human still can recognize an object from incomplete structural edges. We model the problem as a conditional image composition and approach it through a two-stage pipeline. The algorithm first performs a structure analysis to find the structure importance map of the object. Then the importance map is combined into a conditional mean value composition to generate the hidden image. Experimental results show that the proposed algorithm can efficiently generate visually appealing hidden images. Moreover, a user study is conducted to validate the effectiveness of the proposed algorithm.