Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/92768
Title: | 視覺影像評鑑:用於評價攝影中的散景和整體影像質量的模型 Assessing Visual Quality: A Comprehensive Model for Evaluating Bokeh Effects and Overall Quality in Photography |
Authors: | 簡哲元 Che-Yuan Chien |
Advisor: | 廖世偉 Shih-Wei Liao |
Keyword: | 視覺影像評鑑,視覺審美評鑑,散景效果, Image Quality Assessment,Image Aesthetic Assessment,Bokeh Effects, |
Publication Year : | 2024 |
Degree: | 碩士 |
Abstract: | 隨著多媒體和互聯網技術的快速發展,影像質量顯著提升了用戶體驗。本文重點研究散景影像的質量評估,這是一種通過創造模糊背景來突出主要主體的流行攝影技術。儘管散景效果在業餘和專業攝影師中廣泛使用,但對其質量的研究仍然有限。為了填補這一空白,我們開發了一個專門的數據集,系統地評估和比較帶有散景效果的圖像和整體質量。
我們介紹了散景視覺融合網(BSFN),這是一種用於評估圖像美學的模型,並在已建立的數據集上取得了顯著成果。我們的研究包括創建了一個散景圖像評分數據集(BISD),該數據集包含有經驗的攝影師和普通用戶的評分,以及開發了BSFN模型。該模型在BAID數據集上的準確率達到了77.89%,斯皮爾曼等級相關係數(SRCC)和皮爾森相關係數(PCC)分別為0.475和0.533,達到了此類評估目前最高的準確率。 我們希望這項研究能夠為進一步探索散景和美學評估奠定基礎,為提升影像質量評估盡一份心力。 With the rapid advancement of multimedia and internet technology, the quality of images significantly enhances the user experience. This paper focuses on the quality assessment of bokeh images, a popular photography technique used to highlight main subjects by creating a blurred background. Despite its widespread use among both amateur and professional photographers, research on the quality of bokeh images remains limited. To address this gap, we developed a specialized dataset to systematically evaluate and compare the overall quality of images with bokeh effects. We introduce the Bokeh Sight Fusion Net (BSFN), a model designed to assess image aesthetics, which has achieved significant results on established datasets. Our research includes the creation of a Bokeh Image Scoring Dataset (BISD), enriched with ratings from experienced photographers and general users, and the development of the BSFN model. This model has demonstrated an accuracy rate of 77.89% on the BAID dataset, with Spearman’s rank correlation coefficient (SRCC) and Pearson’s correlation coefficient (PCC) of 0.475 and 0.533, respectively, marking the highest accuracy to date for such assessments. We hope this research will provide a foundation for further exploration of bokeh and aesthetic assessments, and contribute to the enhancement of image quality evaluation. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/92768 |
DOI: | 10.6342/NTU202401197 |
Fulltext Rights: | 未授權 |
Appears in Collections: | 資訊工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-112-2.pdf Restricted Access | 29.63 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.