Imperceptible Pattern Embedding: Structured Light Steganography For Augmented Reality Applications

Thumbnail Image
Maquignaz, Ian
PROCAM , Machine Vision , Steganography , Computer Vision , 3D Imaging , Depth Sensing
In this work, we describe a novel PROCAM sensor which uses steganographic techniques to embed imperceptible patterns in structured light, allowing a projector to function both as a versatile display device and a sensor for augmented reality applications. The purpose of this technology is to enable a new generation of intelligent projectors with the capacity for self-calibration, responsive user-interaction and the ability communicate with peer devices. Our approach analyses input frames by transformation into the frequency domain where each frame is encoded with an imperceptible pattern. The encoded input frame is then returned to the spatial domain where it is projected and in turn recaptured by the PROCAM pair. We decode by filtering the image in the frequency domain and extracting the transformed encoded pattern which allows the estimation of the transformation incurred by projection or the extraction encoded data. It is hoped that the methodology described herein will enable future work in PROCAM AR calibration, stenographic communication, and environment sensing AR systems with the capacity responsive user-interfaces, be that through behaviour recognition or other forms of interaction. Our results show that the proposed system is capable of embedding patterns into input frames with only minimal alteration to the input image, and pattern extraction can be achieved reliably with minimal noise.
External DOI