A new multiperspective framework for standardization and benchmarking of image dehazing algorithms
A standardization and benchmarking framework for image dehazing algorithms based on multiple perspectives is not yet available. Hence, this study proposed a new multi-perspective standardization and benchmarking framework for image dehazing algorithms. Experiments were conducted in three main phases...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://eprints.uthm.edu.my/1789/2/KARRAR%20HAMEED%20ABDUL%20KAREEM%20-%20declaration.pdf http://eprints.uthm.edu.my/1789/1/KARRAR%20HAMEED%20ABDUL%20KAREEM%20-%2024p.pdf http://eprints.uthm.edu.my/1789/3/KARRAR%20HAMEED%20ABDUL%20KAREEM%20-%20fulltext.pdf http://eprints.uthm.edu.my/1789/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A standardization and benchmarking framework for image dehazing algorithms based on multiple perspectives is not yet available. Hence, this study proposed a new multi-perspective standardization and benchmarking framework for image dehazing algorithms. Experiments were conducted in three main phases. First, the image dehazing criteria were standardized based on Fuzzy-Delphi Method (FDM). Furthermore, an objective experiment was conducted to test and evaluate the selected criteria from FDM within constraints of Pearson Linear Correlation Coefficient (PLCC) and Spearman Rank Order Correlation Coefficient (SRCC). Second, an evaluation experiment was conducted to obtain a new multi-perspective decision matrix. Third, Best Worst Method (BWM) and Vlse Kriterijumska Optimizacija Kompromisno Resenje (VIKOR) methods were hybridized to determine the weight of the standardized criteria and rank the algorithms. To objectively validate the selection results, mean was applied for this purpose. To evaluate the proposed framework, two main approaches were applied. On the one hand, a standard dataset was tested on the selected criteria and image dehazing algorithms to select the best algorithm. On the other hand, a benchmarking checklist scenario was adopted to measure the feasibility of the proposed work compared to other methods. The results revealed that 11 criteria were selected as the best according to FDM stipulations. Furthermore, seven criteria had been satisfied with the PLCC and SRCC tests. Hybridization of BWM and VIKOR methods can effectively solve the challenges in the selection of the optimal algorithm. The ranking results identified Contrast Limited Adaptive Histogram Equalization (CLAHE) as the best image dehazing algorithm. Apart from that, the benchmarking checklist scenario showed the proposed framework was more effective than the benchmark study. |
|---|