Enhancing Depth Consistency in Augmented and Diminished Reality : Techniques and Evaluations Using RGB Imagery

Augmented Reality (AR) applications are rapidly gaining popularity across various industries, including education and marketing. By integrating real-world environments with virtual objects, AR enhances user understanding and information display for products. This paper explores Diminished Reality (DR) techniques, which are used to visually remove real objects from AR environments. Despite growing interest, much of the DR research predominantly focuses on maintaining consistency between real and virtual elements, particularly in texture handling on marker areas. Our study addresses the preservation of depth consistency using edge detection and planar segmentation to construct a depth map, essential for developing effective DR methods. We introduce a two-stage process involving depth mask construction, each stage equipped with error measurement for iterative refinement. Our proposed techniques, Planarity and Boundary Depth, are evaluated on a dataset of high-quality RGB images captured by digital cameras. Experimental results validate the effectiveness of our methods across various performance metrics, confirming the practicality of our approach in enhancing AR experiences.