dc.creator | Mairgiotis A., Tsampra D., Kondi L.P. | en |
dc.date.accessioned | 2023-01-31T08:55:54Z | |
dc.date.available | 2023-01-31T08:55:54Z | |
dc.date.issued | 2021 | |
dc.identifier | 10.1109/PCS50896.2021.9477453 | |
dc.identifier.isbn | 9781665425452 | |
dc.identifier.uri | http://hdl.handle.net/11615/76119 | |
dc.description.abstract | The adoption of a Natural Scene Statistics (NSS) model has been an important research direction in the selection of perceptual features capable of giving satisfactory results in the problem of image quality assessment (IQA). In this work, trying to improve the performance of a blind IQA methodology, we simultaneously consider quality aware features from both the spatial and the transform domains. Moreover, for the first time, a statistical description of the spatial domain is investigated through the Student's t distribution, trying to predict the subjective evaluation of humans and to reduce the total number of features. In essence, a large number of features are used, which are optimized by the consequent characterization with the distribution's parameters. The proposed model is then fed to a tool to learn a simple regression model. In this way the extracted trained model is used to predict the graded image quality score, based on known publicly available datasets. The results are interesting and show high levels of agreement with the subjective human perception while maintaining a low total number of features. © 2021 IEEE. | en |
dc.language.iso | en | en |
dc.source | 2021 Picture Coding Symposium, PCS 2021 - Proceedings | en |
dc.source.uri | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85112059304&doi=10.1109%2fPCS50896.2021.9477453&partnerID=40&md5=fa257be9d890c680df358ea99c98dea7 | |
dc.subject | Regression analysis | en |
dc.subject | Image quality assessment (IQA) | en |
dc.subject | Natural scene statistics | en |
dc.subject | Perceptual feature | en |
dc.subject | Regression model | en |
dc.subject | Statistical descriptions | en |
dc.subject | Student's t distribution | en |
dc.subject | Subjective evaluations | en |
dc.subject | Transform domain | en |
dc.subject | Image quality | en |
dc.subject | Institute of Electrical and Electronics Engineers Inc. | en |
dc.title | Improved hybrid blind IQA using alternative NSS characterization in the spatial domain | en |
dc.type | conferenceItem | en |