Reproduce “A Wavelet-based Stereo Matching Framework for Solving Frequency Convergence Inconsistency” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xiaobao Wei, Jiawei Liu, Dongbo Yang, Junda Cheng, Changyong Shu, Wei Wang
A new framework for iterative stereo matching, Wavelet-Stereo, addresses frequency convergence inconsistency by explicitly decomposing images into high and low-frequency components using Discrete Wavelet Transform. It achieves top rankings on the KITTI 2015 and KITTI 2012 benchmarks, setting new state-of-the-art performance on Scene Flow and ETH3D datasets.
Key insights
- Iterative stereo matching models exhibit frequency convergence inconsistency, causing high-frequency detail degradation when all frequency components are optimized uniformly.
- Explicitly separating and processing high and low-frequency information with dedicated architectural components (DWT, specialized feature extractors, HPU) is crucial for accurate and robust disparity estimation.
- Adaptive, bidirectional interaction between high and low-frequency features within an iterative update mechanism can effectively preserve fine details while refining global disparity structure.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.