.png){kind=icon}
.png){kind=icon}
Return
Adaptive Depth Completion Optimization Based on Traditional Image Processing
.png){kind=icon}
PRE
AI
DOI：10.1016/j.dsp.2025.105642.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Depth completion is a core technique in computer vision and plays a vital role in applications such as autonomous driving, robot navigation, and 3D reconstruction. To address the limitations of classical image processing methods in adapting to local density variations and preserving edge details, this paper proposes a sequential joint optimization method that integrates adaptive dilation and region-adaptive filtering. The proposed method first adjusts the dilation kernel size dynamically based on local density information to achieve efficient hole filling. Subsequently, image gradients are used to guide region-aware filtering, enabling effective noise suppression while enhancing structural details. These two stages complement each other to jointly improve global accuracy and local fidelity, enhancing both the robustness and precision of depth completion. The proposed method is validated on the KITTI dataset, and experimental results demonstrate significant improvements in completion accuracy and edge preservation compared to classical approaches, offering a more robust optimization solution within classical image processing frameworks.
Journal
D
IF:
3
Papers:
687
Citations:
0