Abstract
A lot of engineering applications related to computer vision use stereo vision algorithms and they usually require fast processing capability near to real time. In this paper, we present a new technique of area-based stereo matching algorithm which provides faster processing capability by using block-based matching. Our algorithm employs block-based matching followed by pixel-based matching. By applying block-based matching hierarchically, representative disparity values are assigned to each block. With these rough results, dense disparity map is acquired under very short search range. This procedure can reduce processing time greatly. We test our matching algorithms for various types of images, and shall show good performance of our stereo matching algorithm in both running time and correctness' aspect.
| Original language | English |
|---|---|
| Pages (from-to) | 789-798 |
| Number of pages | 10 |
| Journal | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
| Volume | 3044 |
| Publication status | Published - 2004 |
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)