Content Adaptive Embedding of Complementary Patterns for Nonintrusive Direct-Projected Augmented Reality

摘要

In direct-projected augmented reality, the visual patterns for compensation may distract users despite users would not be interested in the compensation process. The distraction becomes more serious for dynamic projection surface in which compensation and display should be done simultaneously. Recently, a complementary pattern-based method of efficiently hiding the compensation process from users' view has been proposed. However, the method faced the tradeoff between the pattern imperceptibility and compensation accuracy. In this paper, we embed locally different strength of pattern images into different channels of the projector input images (AR images) after analyzing their spatial variation and color distribution. It is demonstrated that our content adaptive approach can significantly improve the imperceptibility of the patterns and produce better compensation results by comparing it with the previous approach through a variety of experiments and subjective evaluation.