Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Compressed Depth Map Super-Resolution and Restoration: AIM 2024 Challenge Results

Marcos V. Conde, Florin-Alexandru Vasluianu, Jinhui Xiong, Wei Ye, Rakesh Ranjan, Radu Timofte

TL;DR

This challenge introduces a focus on developing innovative depth upsampling techniques to reconstruct high-quality depth maps from compressed data, crucial for overcoming the limitations posed by depth compression.

Abstract

The increasing demand for augmented reality (AR) and virtual reality (VR) applications highlights the need for efficient depth information processing. Depth maps, essential for rendering realistic scenes and supporting advanced functionalities, are typically large and challenging to stream efficiently due to their size. This challenge introduces a focus on developing innovative depth upsampling techniques to reconstruct high-quality depth maps from compressed data. These techniques are crucial for overcoming the limitations posed by depth compression, which often degrades quality, loses scene details and introduces artifacts. By enhancing depth upsampling methods, this challenge aims to improve the efficiency and quality of depth map reconstruction. Our goal is to advance the state-of-the-art in depth processing technologies, thereby enhancing the overall user experience in AR and VR applications.

Compressed Depth Map Super-Resolution and Restoration: AIM 2024 Challenge Results

TL;DR

This challenge introduces a focus on developing innovative depth upsampling techniques to reconstruct high-quality depth maps from compressed data, crucial for overcoming the limitations posed by depth compression.

Abstract

The increasing demand for augmented reality (AR) and virtual reality (VR) applications highlights the need for efficient depth information processing. Depth maps, essential for rendering realistic scenes and supporting advanced functionalities, are typically large and challenging to stream efficiently due to their size. This challenge introduces a focus on developing innovative depth upsampling techniques to reconstruct high-quality depth maps from compressed data. These techniques are crucial for overcoming the limitations posed by depth compression, which often degrades quality, loses scene details and introduces artifacts. By enhancing depth upsampling methods, this challenge aims to improve the efficiency and quality of depth map reconstruction. Our goal is to advance the state-of-the-art in depth processing technologies, thereby enhancing the overall user experience in AR and VR applications.
Paper Structure (20 sections, 4 equations, 11 figures, 1 table)

This paper contains 20 sections, 4 equations, 11 figures, 1 table.

Table of Contents

  1. Introduction
  2. Challenge Dataset
  3. Associated AIM Challenges.
  4. Proposed Methods
  5. General Ideas
  6. A Simple and Effective Baseline for Depth Upsampling and Refinement
  7. Implementation details
  8. DAS-Depth: Depth-Aware Scale Refinement for Monocular Depth Estimation
  9. Implementation details
  10. Attaching ControlNet to an RGB2Depth Model for Depth Upsampling
  11. Method Description
  12. Implementation details
  13. Relative depth guided depth map upsampling
  14. Model design.
  15. RAFT-DU: Depth Upsampling Using RAFT Structures
  16. ...and 5 more sections

Figures (11)

  • Figure 1: A graphical representation of the degradations suffered by a High-Resolution (HR) depth map (A), being mapped to its corresponding Low-Resolution (LR) version. Bitdepth reduction (B), spatial downscaling (C) and characteristic noise are applied to produce the Low-Quality (LQ) compressed depth map (D).
  • Figure 2: Samples from the Testing Phase split, consisting of the HR RGB image (A), the HR reference depth map (B), and the upscaled LR input depth map (C). The participants only have access to the HR RGB image and the LR Depth map.
  • Figure 3: Visual comparisons of the proposed solutions on test samples.
  • Figure 4: The overall framework of the team UM-IT proposal.
  • Figure 5: DAS-Depth analysis of depth maps in the Training Set: (A) Inconsistent depth in Sky Regions; (B) Noise in Low-Resolution depth maps for smaller depth values
  • ...and 6 more figures