Abstract:Although depth cameras acquire depth in dynamic scenes, the captured depth images are often noisy and of low resolution. Depth images have the physical nature of being represented by smooth regions and edges in between them, i.e. depth images have high edge sparsity in the gradient domain. In this paper, we propose intensity guided depth upsampling using edge sparsity and super-weighted L 0 gradient minimization. First, we get mutual structure between intensity and depth using joint mutual structure filtering. Second, we generate an initial depth image using recursive interpolation. Next, we generate weights for L 0 gradient minimization based on gradient and entropy of the intensity image, and upsample the depth image using super-weighted L 0 gradient minimization. In super-weighted L 0 gradient minimization, we combine two main terms: Hybrid data fidelity and weighted L 0 gradient regularization. The hybrid data fidelity term combines both zero-order and first-order data differences to suppress staircase artifacts, while the weighted L 0 gradient regularization term preserves depth structures and removes noise. Finally, we further refine the depth image using adaptive fast weighted median filtering. Experiments on Middlebury and realworld scene datasets verify that the proposed method produces edge preserving depth upsampling results and outperforms state-of-the-arts in terms of both visual quality and quantitative measurements.

Upsampling Range Camera Depth Maps Using High-Resolution Vision Camera and Pixel-Level Confidence Classification.

A Depth Estimation Framework Based on Unsupervised Learning and Cross-Modal Translation

A Comparative Study on Guided Depth Enhancement

Robust High Quality Image Guided Depth Upsampling

Depth Upsampling Method Via Markov Random Fields Without Edge-Misaligned Artifacts

Depth super resolution using bilateral filter

Efficient Confidence-Based Hierarchical Stereo Disparity Upsampling for Noisy Inputs

Depth Map Upsampling via Progressive Manner Based on Probability Maximization.

An Extremely Effective Spatial Pyramid and Pixel Shuffle Upsampling Decoder for Multiscale Monocular Depth Estimation

Temporal Upsampling of Depth Maps Using a Hybrid Camera

Depth map enhancement based on color and depth consistency

Upsampling The Depth Map With Its Own Properties

Depth Estimation from Monocular Images Using Dilated Convolution and Uncertainty Learning.

Depth upsampling algorithm based on image edge characteristics

High-Resolution Depth Maps Based on TOF-Stereo Fusion

Depth Map Upsampling Via Compressive Sensing

URCDC-Depth: Uncertainty Rectified Cross-Distillation with CutFlip for Monocular Depth Estimation

Intensity Guided Depth Upsampling Using Edge Sparsity and Super-Weighted L Gradient Minimization.

Depth Map Super-Resolution Using Stereo-Vision-Assisted Model

Improving Sub-Pixel Correspondence Through Upsampling.