Development of quality assessment methods for wood images / Heshalini Rajagopal @ Ramasamy
Image Quality Assessment (IQA) is a vital element in improving the efficiency of an automatic recognition system of various wood species. There is a need to develop a No- Reference Image Quality Assessment (NR-IQA) system as a perfect and distortion free wood images may be impossible to be acquired...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/13744/2/Heshalini.pdf http://studentsrepo.um.edu.my/13744/1/Heshalini.pdf http://studentsrepo.um.edu.my/13744/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Malaya |
| Summary: | Image Quality Assessment (IQA) is a vital element in improving the efficiency of an automatic recognition system of various wood species. There is a need to develop a No- Reference Image Quality Assessment (NR-IQA) system as a perfect and distortion free wood images may be impossible to be acquired in the dusty and dark environment in timber factories. Many IQAs which focus on some image of interest such as natural images have been proposed. However, an IQA specifically for wood images have not been proposed so far. Hence, this thesis proposes two No-Reference IQA (NR-IQA) metrics, Modified BRISQUE Wood Image Quality Assessment (MBW-IQA) and GLCM and Gabor Wood Image Quality Assessment (GGW-IQA) to assess the quality of wood images. Firstly, Support Vector Machine (SVM) Regression (SVR) was trained using Generalized Gaussian Distribution (GGD) and Asymmetric Generalized Gaussian Distribution (AGGD) features calculated for wood images together with the mean opinion score (MOS) obtained from subjective evaluation to develop the MBW-IQA. Secondly, SVR was trained using Gray Level Co-Occurrence Matrix (GLCM) and Gabor features calculated for wood images together with the MOS to develop the GGW-IQA metric. The MBW-IQA and GGW-IQA metrics are compared with one of the established NR-IQA metrics, namely, Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE), Deep Neural Network IQA (deepIQA), Deep Bilinear Convolution Neural Network (DBCNN) and five Full Reference-IQA (FR-IQA) metrics known as MSSIM, SSIM, FSIM, IWSSIM and GMSD. Results shows that the GGW-IQA outperforms the MBW-IQA, BRISQUE, deepIQA, DB-CNN and all the FR-IQA metrics. Moreover, the GGW-IQA metric is beneficial in wood industry as a distortion free reference image is not needed to evaluate the wood images.
|
|---|