Parallelization of modified geodesic active contour model on high-resolution satellite image for segmentation process
A modified Sign Pressure Force (SPF) function able to enhance the existing Edge Stopping Function (ESF) in terms of simulation, visualization, and segmentation of high-resolution satellite images of Nusajaya using the Geodesic Active Contour (GAC) model. The modified SPF function is formulated by in...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/102677/1/MaizatulNadirahMustaffaPFS2020.pdf.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A modified Sign Pressure Force (SPF) function able to enhance the existing Edge Stopping Function (ESF) in terms of simulation, visualization, and segmentation of high-resolution satellite images of Nusajaya using the Geodesic Active Contour (GAC) model. The modified SPF function is formulated by integrating both the local SPF function and the global SPF function. Next, the modified GAC model is extended to a higher-order modified GAC model. The second-order and fourth-order modified GAC models are implemented using the Finite Difference Method (FDM) and developed into a tri-diagonal and Penta-diagonal Linear System of Equations (LSE). Some numerical methods such as Second-Order Alternating Group Explicit (AGE2), Second-Order Red-Black Gauss-Seidel (RBGS2), and Second-Order Jacobi (JB2) methods are used to solve the LSE of second-order modified GAC model. Meanwhile, Fourth-Order Alternating Group Explicit (AGE4), Fourth-Order Red-Black Gauss-Seidel (RBGS4), and Fourth-Order Jacobi (JB4) methods are used to solve the LSE of the fourth-order modified GAC model. The sequential algorithm is developed using Matlab R2015a software. The indicator of numerical results is analyzed based on execution time, number of iterations, maximum error, root mean square error, and computational complexity. The actual high-resolution satellite images of Nusajaya generate a large amount of data, resulting in an enormous amount of execution time and high computational complexity. Thus, the implementation of a parallel algorithm is a reliable alternative for improving the sequential computation and reduced the execution time up to 82.23%. The parallel computation obtains an extensive large scale simulation capability for high-resolution satellite image data. The domain decomposition strategy is implemented by using the Matlab parallel computing toolbox based on the shared memory architecture. Parallel performance evaluations of numerical methods are measured based on speedup, efficiency, effectiveness, temporal performance, and granularity. As a conclusion, this investigation has proven the second-order modified GAC model could be extended to a fourth-order modified GAC model to simulate and visualize edge-region segmentation of high-resolution satellite images. Consequently, the Parallel Fourth-Order Alternating Group Explicit (PAGE4) method is an alternative solution for large sparse segmentation process of high-resolution satellite images of Nusajaya as it improves the performance up to 82.26%. Based on the numerical results and parallel performance measurements, the parallel algorithm is proved to reduce the execution time and computational complexity up to 82.23% compared to the sequential algorithm. |
|---|