Development of Arabic word pose estimation algorithm /
This study investigates possible combination of matching technique with Infinitesimal Plane-Based Pose Estimation (IPPE) that suits better in estimating the pose of Arabic text images. The pattern matching technique involves are Speeded-Up Robust Features (SURF) and Affine Scale Invariant Feature Tr...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Subjects: | |
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/4585 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study investigates possible combination of matching technique with Infinitesimal Plane-Based Pose Estimation (IPPE) that suits better in estimating the pose of Arabic text images. The pattern matching technique involves are Speeded-Up Robust Features (SURF) and Affine Scale Invariant Feature Transform (ASIFT). The experiment is demonstrated in Arabic word images from different angles of viewpoints. The algorithms are tested on a dataset chosen from a few words within Surah Al-Fatihah in the Quran. A total of 260 images were taken from the left and right side of the image. Then, a set of sub-words were recognized and the performance were tested. Thus, this study focuses on comparing the performance of the technique against Arabic words in two sub-words or one sub-word form. We evaluated the performance by analyzing the matching accuracy rate and how it affects the pose estimation. Based on the results obtained for the pattern matching technique performance on Arabic scripts, the experiment result is used as a guide in estimating a pose of the target images in different sub-words. Generally, ASIFT shows a better accuracy rate than SURF by 11.42 percent. However, after displaying the IPPE and refined IPPE camera pose, SURF shows a better performance as compared to ASIFT. The overall results of the study signify that good IPPE pose does not rely on the accuracy rate of matching inliers with original interest points. The study also demonstrates that one sub-words shows a better accuracy rate than with two sub-words caused by unnecessary interest points detected. |
|---|---|
| Physical Description: | xiii, 58 leaves : illustrations ; 30cm. |
| Bibliography: | Includes bibliographical references (leaves 50-53). |