Development Of Double Stage Filter (DSF) On Stereo Matching Algorithm For 3D Computer Vision Applications
In the field of stereo vision, some of existing stereo matching algorithms are designed with less accuracy of algorithm. Thus, a new hybrid algorithm with higher accuracy of computation is developed in this project. This thesis will present the design, development and analysis of performance on a de...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2016
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/18177/1/Development%20Of%20Double%20Stage%20Filter%20%28DSF%29%20On%20Stereo%20Matching%20Algorithm%20For%203D%20Computer%20Vision%20Applications%2024%20Pages.pdf http://eprints.utem.edu.my/id/eprint/18177/2/Development%20Of%20Double%20Stage%20Filter%20%28DSF%29%20On%20Stereo%20Matching%20Algorithm%20For%203d%20Computer%20Vision%20Applications.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In the field of stereo vision, some of existing stereo matching algorithms are designed with less accuracy of algorithm. Thus, a new hybrid algorithm with higher accuracy of computation is developed in this project. This thesis will present the design, development and analysis of performance on a developed Double Stage Filter (DSF) algorithm and other existing stereo matching algorithms. DSF algorithm is a hybrid stereo matching algorithm which divided into two phases. Phase 1 is consists of the part on Sum of Absolute Differences from basic block matching and the part of Scanline Optimization (SO) from dynamic programming approches while phase 2 includes segmentation, merging and basic median filter process. The main feature of DSF algorithm is mainly on the phase 2 or as post-processing in which to remove the unwanted aspects like random noises and horizontal streaks, which is obtained from the raw disparity depth map on the step of optimization. In order to remove the unwanted aspects, two stages filtering process are needed along with the developed approaches in the phase 2 of DSF algorithm. There are two categorized evaluations done on the disparity maps obtained by the algorithms : objective evaluation and subjective evaluation. The objective evaluation includes the evaluation system of Middlebury Stereo Vision website page, computation analysis and traditional methods of Mean Square Errors (MSE), Peak to Signal Noise Ratio (PSNR) and Structural Similarity Index Metric (SSIM). Besides, for subjective evaluation, the datasets are captured from LNC IP camera and the results obtained by the selected algorithms are evaluated by human's eyes perception. Based on the results of evaluations, the results obtained by DSF is better than the algorithms, basic block matching and dynamic programming. |
|---|