Augmented Reality Marker Hiding with Texture Deformation.

Area of Science:

• Computer Vision
• Computer Graphics
• Human-Computer Interaction

Background:

• Augmented reality (AR) marker hiding aims to seamlessly remove markers from video streams.
• Conventional methods using homography matrices fail on non-planar surfaces, causing texture discontinuities.

Purpose of the Study:

• To propose a novel AR marker hiding technique that overcomes limitations of planar surface assumptions.
• To achieve marker hiding without visual artifacts, even with complex background geometries.

Main Methods:

• Estimates dense background motion by analyzing sparse feature point movements around the marker.
• Applies a smooth motion assumption to deform the background image accurately.
• Avoids explicit measurement of non-planar background geometry.

Main Results:

• Demonstrates effective AR marker hiding across diverse environments with varying geometries and textures.
• Successfully eliminates discontinuities at texture boundaries, enhancing visual integration.
• Achieves seamless visual blending of the background with the marker region.

Conclusions:

• The proposed method offers a robust solution for AR marker hiding on non-planar surfaces.
• It significantly improves the visual quality and immersion in AR applications.
• This technique advances the state-of-the-art in real-time video processing for AR.