Real Time Dense Depth Estimation by Fusing Stereo with Sparse Depth\n Measurements

2018/09/20 by Shreyas S. Shivakumar, Shivakumar, Shreyas S., Kartik Mohta +7 · 1 citation
Computer Science · Engineering · #Advanced Vision and Imaging #Image Processing Techniques and Applications #Optical measurement and interference techniques

paper · pdf · doi:10.48550/arxiv.1809.07677

Abstract

We present an approach to depth estimation that fuses information from a\nstereo pair with sparse range measurements derived from a LIDAR sensor or a\nrange camera. The goal of this work is to exploit the complementary strengths\nof the two sensor modalities, the accurate but sparse range measurements and\nthe ambiguous but dense stereo information. These two sources are effectively\nand efficiently fused by combining ideas from anisotropic diffusion and\nsemi-global matching.\n We evaluate our approach on the KITTI 2015 and Middlebury 2014 datasets,\nusing randomly sampled ground truth range measurements as our sparse depth\ninput. We achieve significant performance improvements with a small fraction of\nrange measurements on both datasets. We also provide qualitative results from\nour platform using the PMDTec Monstar sensor. Our entire pipeline runs on an\nNVIDIA TX-2 platform at 5Hz on 1280x1024 stereo images with 128 disparity\nlevels.\n

Cited by

• Diffusion Knows Transparency: Repurposing Video Diffusion for Transparent Object Depth and Normal Estimation

Related

• Sparse-to-Dense: Depth Prediction from Sparse Depth Samples and a Single Image
• Stereo-LiDAR Depth Estimation with Deformable Propagation and Learned Disparity-Depth Conversion
• Balanced Depth Completion between Dense Depth Inference and Sparse Range Measurements via KISS-GP
• LiStereo: Generate Dense Depth Maps from LIDAR and Stereo Imagery
• Sparse LiDAR and Stereo Fusion (SLS-Fusion) for Depth Estimationand 3D Object Detection