Absorbing Markov chain saliency based variational selective active contour model for grayscale and vector-valued images / Muhammad Syukri Mazlin
Image segmentation plays a crucial role in various fields, including medical image analysis, pattern recognition, computer vision and biometric identification where accurate delineation of desired objects by the image segmentation model is essential. The well-known image segmentation model namely va...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/106803/1/106803.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Image segmentation plays a crucial role in various fields, including medical image analysis, pattern recognition, computer vision and biometric identification where accurate delineation of desired objects by the image segmentation model is essential. The well-known image segmentation model namely variational Active Contour (AC) model is capable in partitioning or segmenting objects in an input image effectively. To partition or segment a specific object in an input image, the selective type of AC model segmentation approach is preferable as compared to the global type of AC model. However, when it comes to segment grayscale or vector-valued (color) images with inhomogeneous intensity, the existing selective AC models often yield unsatisfactory results. To address this issue, two studies namely Study 1 and Study 2 are designed in this research which aim to propose a new variational selective AC model for grayscale and vector-valued images respectively by integrating the saliency image map and the local image fitting concepts. By leveraging the saliency image map via Absorbing Markov Chain approach, the model enhances the focus on desired objects while suppressing the influence of intensity inhomogeneity. In addition, the presence of the local image fitting helps in dealing with intensity inhomogeneity problem by effectively capturing local image characteristics. |
|---|