Abstract:The edge information plays a key role in the restoration of a depth map. Most conventional methods assume that the color image and depth map are consistent in edge areas. However, complex texture regions in the color image do not match exactly with edges in the depth map. In this paper, firstly, we point out that in most cases the consistency between normal map and depth map is much higher than that between RGB-D pairs. Then we propose a dual normal-depth regularization term to guide the restoration of depth map, which constrains the edge consistency between normal map and depth map back and forth. Moreover, considering the bimodal characteristic of weight distribution that exists in depth discontinuous areas, a reweighted graph Laplacian regularizer is proposed to promote this bimodal characteristic. And this regularization is incorporated into a unified optimization framework to effectively protect the piece-wise smoothness(PWS) characteristics of depth map. By treating depth image as graph signal, the weight between two nodes is adapted according to its content. The proposed method is tested for both noise-free and noisy cases, and is compared against the state-of-the-art methods on both synthesis and real captured datasets. Extensive experimental results demonstrate the superior performance of our method compared with most state-of-the-art works in terms of both objective and subjective quality evaluations. Specifically, our method is more effective on edge areas and more robust to noises.

Depth dithering based on texture edge-assisted classification

Spatial-temporal depth de-noising for Kinect based on texture edge-assisted depth classification

Guided depth enhancement via a fast marching method

Depth map enhancement based on color and depth consistency

Optimized dithering technique in frequency domain for high-quality three-dimensional depth data acquisition

A Comparative Study on Guided Depth Enhancement

Depth Map Inpainting Using a Fully Convolutional Network

Depth Map Optimization Based on Just Noticeable Depth Difference Model

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

Depth Map Super-Resolution Based on Edge-Guided Joint Trilateral Upsampling

Self-Distilled Depth Refinement with Noisy Poisson Fusion

An Asymmetric Edge Adaptive Filter for Depth Generation and Hole Filling in 3DTV

Hierarchical Joint Bilateral Filtering for Depth Post-Processing

A weighted least squares algorithm for time-of-flight depth image denoising

Ddrnet: Depth Map Denoising And Refinement For Consumer Depth Cameras Using Cascaded Cnns

Reconstructing Sparsely Represented High-resolution Depth Map from Corrupted Samples

Depth Map Denoising Using Graph-Based Transform and Group Sparsity

Depth Map Super-resolution Based on Dual Normal-depth Regularization and Graph Laplacian Prior

Iterative Error Removal for Time-of-Flight Depth Imaging

Depth Super-Resolution Using Joint Adaptive Weighted Least Squares and Patching Gradient.

High quality depth maps from stereo matching and ToF camera.