Reduced-Reference High Dynamic Range Image Quality Asessment Based on Multi Exposure Fusion Algorithm
Total View This Week0
Total View This Week0
Institusion
Universitas Bakrie
Author
Cloramidina, Ocarina
Subject
Computer Science
Datestamp
2019-08-29 06:23:39
Abstract :
This thesis presents objective image quality measurements for High Dynamic Range (HDR) images without complete reference information based on the Multi Exposure Fusion (MEF) Algorithm. It focuses on developing HDR Reduced-Reference (RR) image quality models, and especially on investigating overhead data that can improve the predictive ability of the model. Overhead data used previously have been processed by the MEF algorithm as a basis for this research. In addition, variations and combinations of edge strength and lighting features are extracted from the original image and its quality is measured by a complete reference model modified for the RR model for HDR images. Some of them aim to find out whether the complete reference matrix can be modified for the reduced reference model. These features are then combined to get a single value, which corresponds to the predicted subjective score. This method will be evaluated using variations in Quality Evaluation 1 (QE1), Quality Evaluation 2 (QE2), and Quality Evaluation 3 (QE3). The results show that the proposed average edge strength feature extraction method and the average luminance feature are the best methods where QE3 is close to 1.
This thesis presents objective image quality measurements for High Dynamic Range (HDR) images without complete reference information based on the Multi Exposure Fusion (MEF) Algorithm. It focuses on developing HDR Reduced-Reference (RR) image quality models, and especially on investigating overhead data that can improve the predictive ability of the model. Overhead data used previously have been processed by the MEF algorithm as a basis for this research. In addition, variations and combinations of edge strength and lighting features are extracted from the original image and its quality is measured by a complete reference model modified for the RR model for HDR images. Some of them aim to find out whether the complete reference matrix can be modified for the reduced reference model. These features are then combined to get a single value, which corresponds to the predicted subjective score. This method will be evaluated using variations in Quality Evaluation 1 (QE1), Quality Evaluation 2 (QE2), and Quality Evaluation 3 (QE3). The results show that the proposed average edge strength feature extraction method and the average luminance feature are the best methods where QE3 is close to 1.
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
