Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Guided Depth Super-Resolution by Deep Anisotropic Diffusion

About

Performing super-resolution of a depth image using the guidance from an RGB image is a problem that concerns several fields, such as robotics, medical imaging, and remote sensing. While deep learning methods have achieved good results in this problem, recent work highlighted the value of combining modern methods with more formal frameworks. In this work, we propose a novel approach which combines guided anisotropic diffusion with a deep convolutional network and advances the state of the art for guided depth super-resolution. The edge transferring/enhancing properties of the diffusion are boosted by the contextual reasoning capabilities of modern networks, and a strict adjustment step guarantees perfect adherence to the source image. We achieve unprecedented results in three commonly used benchmarks for guided depth super-resolution. The performance gain compared to other methods is the largest at larger scales, such as x32 scaling. Code (https://github.com/prs-eth/Diffusion-Super-Resolution) for the proposed method is available to promote reproducibility of our results.

Nando Metzger, Rodrigo Caye Daudt, Konrad Schindler• 2022

Related benchmarks

TaskDatasetResultRank
Depth Super-ResolutionNYU v2 (test)--
136
Depth Super-ResolutionMiddlebury 2005-2014 (test)
MSE2.52
34
Depth Map Super-ResolutionNYU v2 (test)
Value Errors2.4
28
Depth Super-ResolutionDIML Microsoft Kinect (test)
MSE4.4
24
Depth Super-ResolutionNYU Bicubic downsampling synthetic v2 (test)
RMSE (x4)1.54
20
Depth Super-ResolutionLu Bicubic downsampling synthetic (test)
RMSE (x4)0.96
20
Depth Super-ResolutionMiddlebury Bicubic downsampling synthetic (test)
RMSE (x4)1.2
20
Depth Super-ResolutionRGB-D-D Bicubic downsampling synthetic (test)
RMSE (4x)1.2
19
Guided Depth Super-resolutionNYU V2
RMSE (4x)1.54
12
Guided Depth Super-resolutionLu
RMSE (x4)0.96
11
Showing 10 of 16 rows

Other info

Code

Follow for update