以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：29

、訪客IP：3.144.193.129

姓名	江睿修(Jui-Hsiu Chiang)  查詢紙本館藏	畢業系所	通訊工程學系
論文名稱	以弱監督式學習輔助無參考畫面之拼接全景影像品質評估 (Weakly Supervised Learning Aided No-reference Stitched Panoramic Image Quality Assessment)
相關論文	★ 應用於車內視訊之光線適應性視訊壓縮編碼器設計 ★ 以粒子濾波法為基礎之改良式頭部追蹤系統 ★ 應用於空間與CGS可調性視訊編碼器之快速模式決策演算法 ★ 應用於人臉表情辨識之強健式主動外觀模型搜尋演算法 ★ 結合Epipolar Geometry為基礎之視角間預測與快速畫面間預測方向決策之多視角視訊編碼 ★ 基於改良式可信度傳遞於同質區域之立體視覺匹配演算法 ★ 以階層式Boosting演算法為基礎之棒球軌跡辨識 ★ 多視角視訊編碼之快速參考畫面方向決策 ★ 以線上統計為基礎應用於CGS可調式編碼器之快速模式決策 ★ 適用於唇形辨識之改良式主動形狀模型匹配演算法 ★ 以運動補償模型為基礎之移動式平台物件追蹤 ★ 基於匹配代價之非對稱式立體匹配遮蔽偵測 ★ 以動量為基礎之快速多視角視訊編碼模式決策 ★ 應用於地點影像辨識之快速局部L-SVMs群體分類器 ★ 以高品質合成視角為導向之快速深度視訊編碼模式決策 ★ 以運動補償模型為基礎之移動式相機多物件追蹤
檔案	[Endnote RIS 格式]    [Bibtex 格式]    [相關文章]   [文章引用]   [完整記錄]   [館藏目錄]   至系統瀏覽論文 (2025-8-1以後開放)
摘要(中)	現今沉浸式媒體發展越來越流行，包括 VR、MR等，這些應用的資料源都須經過影像拼接，影像拼接的成果將直接影響使用者觀看體驗，因此拼接影像的品質評估可使拼接時能更有效的獲取拼接效能。關於採用深度學習的拼接全景影像品質評估，現有的公開資料集中沒有大型人工標註數據集，蒐集資料也需要很高的成本，且無參考畫面拼接影像品質評估 (Blind Stitched Image Quality Assessment, BSIQA) 較為符合實際應用。因此，本論文提出以弱監督式學習 (weakly supervised learning) 進行失真偵測 (distortion detection)，其可在資料量少的狀況下取得更多拼接全景影像的拼接失真特徵，以提升整體品質評估效能。此外，針對拼接場景影像以場景資料集作為品質評估下游任務的預訓練資料集，增加拼接場景影像特徵抽取能力。最後，本論文也進行模型壓縮，在品質評估效能提升的同時，將網路以人工設計方式進行壓縮並重新訓練，使得網路模型可以在嵌入式系統中進行即時的品質評估運算。本論文所提之模型壓縮後方案與現有最好方案 DLNR-SIQA 比較，於 ISIQA 資料集的 Spearman 排序相關係數 (Spearman Rank Order Correlation Coefficient, SROCC) ，比 DLNR-SIQA 高 0.0226，Pearson 相關係數 (Pearson Linear Correlation Coefficient, PLCC) 高 0.0149，正規化均方根誤差 (Normalized Root Mean Square Error, NRMSE) 低 0.2566，因此在時間複雜度及評估準確度皆優於現有其他方案。
摘要(英)	Nowadays, the immersive media is more and more popular, including VR, MR, etc. The data sources of these applications must be image stitched. Image stitching result will directly affect the user’s viewing experiment. Therefore, the quality assessment of stitched image can make stitching performance effectively to obtain. About Nowadays, the immersive media is more and more popular, including VR, MR, etc. The data sources of these applications must be applied image stitching. Since the quality of stitched images directly affects users’ viewing experiences, the quality assessment of stitched image can contribute to the performance of image stitching. Regarding the stitched panoramic image quality assessment using deep learning, there are no public large-scale human annotated datasets. Data collection also requires high costs. In practical, Blind Stitched Image Quality Assessment (BSIQA) is more suitable for real-world applications. Hence, this thesis proposes a distortion detector using weakly supervised learning. It can obtain more stitching features of stitched panoramic images with a small amount of training samples, and it can improve the overall performance of quality assessment. In addition, for the stitched scene images, the scene dataset is used as the pretraining dataset for the downstream task of quality assessment, which can improve extracted features of stitched scene images. Finally, this thesis also performs model compression. The model is manually designed to be compressed and retrained while the accuracy of image quality assessment is kept. Accordingly, real-time quality assessment in embedded systems can be achieved. Compared with the state-of-the-art scheme DLNR-SIQA on the ISIQA dataset, the proposed scheme outperforms DLNR-SIQA on the Spearman Rank Order Correlation Coefficient (SROCC) by 0.0226, Pearson Linear Correlation Coefficient (PLCC) by 0.0149, and the Normalized Root Mean Square Error (NRMSE) by -0.2566. To sum up, the time complexity and evaluation accuracy of the proposed scheme are better than those of existing schemes.
關鍵字(中)	★ 拼接全景影像 ★ 無參考畫面 ★ 影像品質評估 ★ 弱監督式學習 ★ 模型壓縮	關鍵字(英)	★ Stitched panoramic image ★ no-reference image ★ image quality assessment ★ weakly supervised learning ★ model compression
論文目次	摘要 I Abstract II 誌謝 IV 目錄 V 圖目錄 VII 表目錄 VIII 第一章 緒論 1 1.1前言 1 1.2研究動機 1 1.3研究方法 2 1.4論文架構 3 第二章 基於深度學習之拼接全景影像的客觀品質評估 4 2.1全景影像拼接之失真類別 4 2.2拼接影像品質評估之技術現況 7 2.3基於監督式學習之拼接影像品質評估技術現況 8 2.4總結 10 第三章 基於弱監督式學習之視覺表示 11 3.1弱監督式學習定義 11 3.2不準確監督 12 3.3總結 14 第四章 本論文所提以弱監督式學習輔助無參考畫面之拼接全景影像品質評估方案 15 4.1系統架構 15 4.2本論文提出之無參考畫面之全景拼接影像客觀品質評估訓練方案 17 4.2.1本論文提出之針對拼接全景影像預訓練 (Phase#1-1) 17 4.2.2本論文提出之基於弱監督式學習 (Weakly Supervised Learning) 之失真分類 (Phase#1-2) 17 4.2.3本論文提出之拼接全景影像評估方案 (Phase#2) 19 4.3本論文採用之模型壓縮方法 21 4.4總結 22 第五章 實驗結果與分析 23 5.1實驗環境與參數 23 5.1.1資料集於訓練及測試階段使用與資料處理 23 5.1.2 訓練階段超參數與硬體環境設定 26 5.1.3 測試階段硬體環境設定 30 5.2本論文所提方案之實驗結果 30 5.2.1 Phase#1-1 與 Phase#1-2 測試結果 30 5.2.2 Phase#2 品質評估準確率 32 5.2.3 Phase#2 測試結果與數據分析 34 5.2.4 模型參數量與時間複雜度比較 36 5.3總結 37 第六章 結論與未來展望 38 參考文獻 39 著作 42 符號表 43
參考文獻	[1] J. Hou, W. Lin, and B. Zhao, “Content-dependency Reduction with Multi-task Learning in Blind Stitched Panoramic Image Quality Assessment,” in Proc. IEEE International Conference on Image Processing (ICIP), pp. 3463-3467, Oct. 2020. [2] H. Ullah, M. Irfan, K. Han, and J. W. Lee, “DLNR-SIQA: Deep Learning-Based No-Reference Stitched Image Quality Assessment,” Sensors, Vol. 20, No. 22, pp. 6457, Nov. 2020. [3] L. Jing and Y. Tian “Self-supervised Visual Feature Learning with Deep Neural Networks: A Survey,” IEEE Transactions on pattern Analysis and Machine Intelligence, Vol. 43, no, 11, pp. 4037-4058, May 2020. [4] S. Gidaris, A. Bursuc, N. Komodakis, P. Pérez, and M. Cord, “Boosting Few-shot Visual Learning with Self-supervision,” in Proc. IEEE/CVF International Conference on Computer Vision, pp. 8059-8068, Oct. 2019. [5] J.-H. Chiang and C.-W. Tang “Self-supervised Learning Aided Blind Stitched Panoramic Image Quality Assessment,” in Proc. IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW), pp. 1-2, Sep. 2021. [6] M. Brown and D. G. Lowe “Automatic Panoramic Image Stitching Using Invariant Features,” International journal of computer vision, Vol. 74, No. 1, pp. 59-73, Dec. 2007. [7] M. A. Fischler, and R. C. Bolles, “Random Aample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography,” Communications of the ACM, Vol. 24, No. 6, pp. 381-395, June 1981. [8] P. C. Madhusudana and R. Soundararajan “Subjective and Objective Quality Assessment of Stitched Images for Virtual Reality,” IEEE Transactions on Image Processing, Vol. 28, No. 11, pp. 5620-5635, June 2019. [9] K. Yu, J. Li, Y. Zhang, Y. Zhao, and L. Xu, “Image Quality Assessment for Omnidirectional Cross-reference Stitching,” arXiv preprint arXiv:1904.04960, Apr. 2019. [10] J. Li, Y. Zhao, W. Ye, K. Yu, and S. Ge, “Attentive Deep Stitching and Quality Assessment for 360◦ Omnidirectional Images,” IEEE Journal of Selected Topics in Signal Processing, Vol. 14, No. 1, pp.209-221, Nov. 2019. [11] Z. H. Zhou, “A Brief Introduction to Weakly Supervised Learning,” National science review, Vol. 5, No. 1, pp. 44-53, Jan. 2018. [11] B. Settles, “Active Learning Literature Survey,” Tech. rep., Technical reports, University of California, Santa Cruz., Vol. 42, 2007. [12] X. Zhu, “Semi-supervised Learning Literature Survey,” 2005. [13] Z. H. Zhou, “Ensemble Methods: Foundations and Algorithms [book review],” IEEE Computational Intelligence Magazine, Vol. 8, No. 1, pp. 77-79, Jan. 2013. [14] P. F. Felzenszwalb, et al, “Object Detection with Discriminatively Trained Part-based Models,” IEEE Transactions on pattern Analysis and Machine Intelligence, Vol. 32, No. 9, pp.1627-1645, Sep. 2010. [15] D. Wang and X. Tan “Label-denoising Auto-encoder for Classification with Inaccurate Supervision Information,” in Proc. IEEE International Conference on Pattern Recognition, pp. 3648-3653, Dec. 2014. [16] D. Hao, L. Zhang, J. Sumkin, A. Mohamed, and S.Wu, “Inaccurate Labels in Weakly-supervised Deep Learning: Automatic Identification and Correction and Their Impact on Classification Performance,” IEEE journal of biomedical and health informatics, Vol. 24, No. 9, pp. 2701-2710, Sep. 2020. [17] J. Xiao, J. Hays, K. Ehinger, A. Oliva, and A. Torralba, “Sun database: Large-scale Scene Recognition from Abbey to Zoo,” in Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 3485-3492, 2010. [18] J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. “ImageNet: A Large-Scale Hierarchical Image Database,” in Proc. IEEE Computer Vision on Pattern Recognition, pp. 248-255, June 2009. [19] M. Huang, Q. Shen, Z. Ma, A. C. Bovik, P. Gupta, R. Zhou, and X.Cao, “Modeling the Perceptual Quality of Immersive Images Rendered on Head Mounted Displays: Resolution and Compression,” IEEE Transactions on Image Processing, Vol. 27, No. 12, pp. 6039-6050, Dec. 2018. [20] A. G. Howard, M. Zhu, B. Chen,D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, and H. Adam, “Mobilenets: Efficient Convolutional Neural Networks for Mobile Vision Applications,” arXiv preprint arXiv:1704.04861, Apr. 2017. [21] M. Tan and Q. V. Le “Efficientnet: Rethinking Model Scaling for Convolutional Neural Networks,” International Conference on Machine Learning, Vol. 97, pp. 6105-6114, 2019.
指導教授	唐之瑋(Chih-Wei Tang)	審核日期	2022-7-15
推文	facebook   plurk   twitter   funp   google   live   udn   HD   myshare   reddit   netvibes   friend   youpush   delicious   baidu
網路書籤	Google bookmarks   del.icio.us   hemidemi   myshare